具备反事实因果推理能力的自主智能体

Accurate trajectory prediction is crucial in autonomous driving to ensure safe and efficient navigation, yet effectively modeling complex interactions among multiple agents remains a significant challenge. Many existing methods still suffer from over-reliance on HD maps, high computational cost, and a lack of interpretability in interaction reasoning. In response, the proposed model innovatively incorporates counterfactual reasoning into social interaction modeling to tackle the challenges of interaction-aware multi-agent trajectory prediction, prioritizing both accuracy and efficiency. In light of the spatiotemporal interaction mechanism and the inherent human cognition governing agents’ motion, the approach simultaneously generates multi-modal trajectories for all agents in a scenario, providing a novel perspective for modeling social interactions through causal reasoning. The results demonstrate that our map-free, lightweight trajectory prediction model rivals the performance of state-of-the-art methods and shows notable improvements over various baselines on publicly available real-world datasets.

Large language model (LLM)-powered agents can translate high-level user intents into plans and actions in an environment. Yet after observing an outcome, users may wonder: What if I had phrased my intent differently? We introduce a framework that enables such counterfactual reasoning in agentic LLM-driven control scenarios, while providing formal reliability guarantees. Our approach models the closed-loop interaction between a user, an LLM-based agent, and an environment as a structural causal model (SCM), and leverages test-time scaling to generate multiple candidate counterfactual outcomes via probabilistic abduction. Through an offline calibration phase, the proposed conformal counterfactual generation (CCG) yields sets of counterfactual outcomes that are guaranteed to contain the true counterfactual outcome with high probability. We showcase the performance of CCG on a wireless network control use case, demonstrating significant advantages compared to naive re-execution baselines.

As humans come to rely on autonomous systems more, ensuring the transparency of such systems is important to their continued adoption. Explainable Artificial Intelligence (XAI) aims to reduce confusion and foster trust in systems by providing explanations of agent behavior. Partially observable Markov decision processes (POMDPs) provide a flexible framework capable of reasoning over transition and state uncertainty, while also being amenable to explanation. This work investigates the use of user-provided counterfactuals to generate contrastive explanations of POMDP policies. Feature expectations are used as a means of contrasting the performance of these policies. We demonstrate our approach in a Search and Rescue (SAR) setting. We analyze and discuss the associated challenges through two case studies.

. We present soid , a tool for interrogating the decision making of autonomous agents using SMT-based automated reasoning. Relying on the Z3 SMT solver and KLEE symbolic execution engine, soid allows investigators to receive rigorously proven answers to factual and counterfactual queries about agent behavior, enabling effective legal and engineering accountability for harmful or otherwise incorrect decisions. We evaluate soid qualitatively and quantitatively on a pair of examples, i) a buggy implementation of a classic decision tree inference benchmark from the explainable AI (XAI) literature; and ii) a car crash in a simulated physics environment. For the latter, we also contribute the soid - gui , a domain-specific, web-based example interface for legal and other practitioners to specify factual and counterfactual queries without requiring sophisticated programming or formal methods expertise.

Hallucination continues to pose a major obstacle in the reasoning capabilities of large language models (LLMs). Although the Multi-Agent Debate (MAD) paradigm offers a promising solution by promoting consensus among multiple agents to enhance reliability, it relies on the unrealistic assumption that all debaters are rational and reflective, which is a condition that may not hold when agents themselves are prone to hallucinations. To address this gap, we introduce the Multi-agent Undercover Gaming (MUG) protocol, inspired by social deduction games like"Who is Undercover?". MUG reframes MAD as a process of detecting"undercover"agents (those suffering from hallucinations) by employing multimodal counterfactual tests. Specifically, we modify reference images to introduce counterfactual evidence and observe whether agents can accurately identify these changes, providing ground-truth for identifying hallucinating agents and enabling robust, crowd-powered multimodal reasoning. MUG advances MAD protocols along three key dimensions: (1) enabling factual verification beyond statistical consensus through counterfactual testing; (2) introducing cross-evidence reasoning via dynamically modified evidence sources instead of relying on static inputs; and (3) fostering active reasoning, where agents engage in probing discussions rather than passively answering questions. Collectively, these innovations offer a more reliable and effective framework for multimodal reasoning in LLMs. The source code can be accessed at https://github.com/YongLD/MUG.git.

While modern recommender systems are instrumental in navigating information abundance, they remain fundamentally limited by static user modeling and reactive decision-making paradigms. Current large language model (LLM)-based agents inherit these shortcomings through their overreliance on heuristic pattern matching, yielding recommendations prone to shallow correlation bias, limited causal inference, and brittleness in sparse-data scenarios. We introduce STARec, a slow-thinking augmented agent framework that endows recommender systems with autonomous deliberative reasoning capabilities. Each user is modeled as an agent with parallel cognitions: fast response for immediate interactions and slow reasoning that performs chain-of-thought rationales. To cultivate intrinsic slow thinking, we develop anchored reinforcement training-a two-stage paradigm combining structured knowledge distillation from advanced reasoning models with preference-aligned reward shaping. This hybrid approach scaffolds agents in acquiring foundational capabilities (preference summarization, rationale generation) while enabling dynamic policy adaptation through simulated feedback loops. Experiments on MovieLens 1M and Amazon CDs benchmarks demonstrate that STARec achieves substantial performance gains compared with state-of-the-art baselines, despite using only 0.4% of the full training data.

Explaining multi-agent systems (MAS) is urgent as these systems become increasingly prevalent in various applications. Previous work has provided explanations for the actions or states of agents, yet falls short in understanding the blackboxed agent’s importance within a MAS and the overall team strategy. To bridge this gap, we propose EMAI, a novel agent-level explanation approach that evaluates the individual agent’s importance. Inspired by counterfactual reasoning, a larger change in reward caused by the randomized action of agent indicates its higher importance. We model it as a MARL problem to capture interactions across agents. Utilizing counterfactual reasoning, EMAI learns the masking agents to identify important agents. Specifically, we define the optimization function to minimize the reward difference before and after action randomization and introduce sparsity constraints to encourage the exploration of more action randomization of agents during training. The experimental results in seven multi-agent tasks demonstrate that EMAI achieves higher fidelity in explanations compared to baselines and provides more effective guidance in practical applications concerning understanding policies, launching attacks, and patching policies.

Hallucination continues to pose a major obstacle in the reasoning capabilities of large language models (LLMs). Although the Multi-Agent Debate (MAD) paradigm offers a promising solution by promoting consensus among multiple agents to enhance reliability, it relies on the unrealistic assumption that all debaters are rational and reflective, which is a condition that may not hold when agents themselves are prone to hallucinations. To address this gap, we introduce the Multi-agent Undercover Gaming (MUG) protocol, inspired by social deduction games like ''Who is Undercover?''. MUG reframes MAD as a process of detecting ''undercover'' agents (those suffering from hallucinations) by employing multimodal counterfactual tests. Specifically, we modify reference images to introduce counterfactual evidence and observe whether agents can accurately identify these changes, providing ground-truth for identifying hallucinating agents and enabling robust, crowd-powered multimodal reasoning. MUG advances MAD protocols along three key dimensions: (1) enabling factual verification beyond statistical consensus through counterfactual testing; (2) introducing cross-evidence reasoning via dynamically modified evidence sources instead of relying on static inputs; and (3) fostering active reasoning, where agents engage in probing discussions rather than passively answering questions. Collectively, these innovations offer a more reliable and effective framework for multimodal reasoning in LLMs.

Pedestrian trajectory prediction is crucial for autonomous systems navigating human-populated environments. However, existing methods face fundamental challenges including spurious correlations induced by confounding social environments, passive uncertainty modeling that limits prediction diversity, and bias coupling during feature interaction that contaminates trajectory representations. To address these issues, we propose a novel Causal Intervention and Counterfactual Reasoning (CICR) framework that shifts trajectory prediction from associative learning to a causal inference paradigm. Our approach features a hierarchical architecture having three core components: a Multisource Encoder that extracts comprehensive spatio-temporal and social context features; a Causal Intervention Fusion Module that eliminates confounding bias through the front-door criterion and cross-attention mechanisms; and a Counterfactual Reasoning Decoder that proactively generates diverse future trajectories by simulating hypothetical scenarios. Extensive experiments on the ETH/UCY, SDD, and AVD datasets demonstrate superior performance, achieving an average ADE/FDE of 0.17/0.24 on ETH/UCY and 7.13/10.29 on SDD, with particular advantages in long-term prediction and cross-domain generalization.

Learning to communicate in order to share state information is an active problem in the area of multi-agent reinforcement learning. The credit assignment problem, the non-stationarity of the communication environment and the problem of encouraging the agents to be influenced by incoming messages are major challenges within this research field which need to be overcome in order to learn a valid communication protocol. This paper introduces the novel multi-agent counterfactual communication learning (MACC) method which adapts counterfactual reasoning in order to overcome the credit assignment problem for communicating agents. Next, the non-stationarity of the communication environment, while learning the communication Q-function, is overcome by creating the communication Q-function using the action policy of the other agents and the Q-function of the action environment. As the exact method to create the communication Q-function can be computationally intensive for a large number of agents, two approximation methods are proposed. Additionally, a social loss function is introduced in order to create influenceable agents, which is required to learn a valid communication protocol. Our experiments show that MACC is able to outperform the state-of-the-art baselines in four different scenarios in the particle environment. Finally, we demonstrate the scalability of MACC in a matrix environment.

Causal world models are systems that can answer counterfactual questions about an environment of interest, i.e. predict how it would have evolved if an arbitrary subset of events had been realized differently. It requires understanding the underlying causes behind chains of events and conducting causal inference for arbitrary unseen distributions. So far, this task eludes foundation models, notably large language models (LLMs), which do not have demonstrated causal reasoning capabilities beyond the memorization of existing causal relationships. Furthermore, evaluating counterfactuals in real-world applications is challenging since only the factual world is observed, limiting evaluation to synthetic datasets. We address these problems by explicitly extracting and modeling causal relationships and propose the Causal Cartographer framework. First, we introduce a graph retrieval-augmented generation agent tasked to retrieve causal relationships from data. This approach allows us to construct a large network of real-world causal relationships that can serve as a repository of causal knowledge and build real-world counterfactuals. In addition, we create a counterfactual reasoning agent constrained by causal relationships to perform reliable step-by-step causal inference. We show that our approach can extract causal knowledge and improve the robustness of LLMs for causal reasoning tasks while reducing inference costs and spurious correlations.

No abstract available

Autonomous multi-agent systems (MAS) are useful for automating complex tasks but raise trust concerns due to risks such as miscoordination or goal misalignment. Explainability is vital for users'trust calibration, but explainable MAS face challenges due to complex environments, the human factor, and non-standardised evaluation. Leveraging the counterfactual effect size model and LLMs, we propose Agentic eXplanations via Interrogative Simulation (AXIS). AXIS generates human-centred action explanations for multi-agent policies by having an LLM interrogate an environment simulator using prompts like'whatif'and'remove'to observe and synthesise counterfactual information over multiple rounds. We evaluate AXIS on autonomous driving across ten scenarios for five LLMs with a comprehensive methodology combining robustness, subjective preference, correctness, and goal/action prediction with an external LLM as evaluator. Compared to baselines, AXIS improves perceived explanation correctness by at least 7.7% across all models and goal prediction accuracy by 23% for four models, with comparable action prediction accuracy, achieving the highest scores overall. Our code is open-sourced at https://github.com/gyevnarb/axis.

We present a methodology for estimating collision risk from counterfactual simulated scenarios built on sensor data from automated driving systems (ADS) or naturalistic driving databases. Two-agent conflicts are assessed by detecting and classifying conflict type, identifying the agents' roles (initiator or responder), identifying the point of reaction of the responder, and modeling their human behavioral expectations as probabilistic counterfactual trajectories. The states are used to compute velocity differentials at collision, which when combined with crash models, estimates severity of loss in terms of probabilistic injury or property damage, henceforth called fractional collisions. The probabilistic models may also be extended to include other uncertainties associated with the simulation, features, and agents. We verify the effectiveness of the methodology in a synthetic simulation environment using reconstructed trajectories from 300+ collision and near-collision scenes sourced from VTTI's SHRP2 database and Nexar dashboard camera data. Our methodology predicted fractional collisions within 1% of ground truth collisions. We then evaluate agent-initiated collision risk of an arbitrary ADS software release by replacing the naturalistic responder in these synthetic reconstructions with an ADS simulator and comparing the outcome to human-response outcomes. Our ADS reduced naturalistic collisions by 4x and fractional collision risk by ~62%. The framework's utility is also demonstrated on 250k miles of proprietary, open-loop sensor data collected on ADS test vehicles, re-simulated with an arbitrary ADS software release. The ADS initiated conflicts that caused 0.4 injury-causing and 1.7 property-damaging fractional collisions, and the ADS improved collision risk in 96% of the agent-initiated conflicts.

Recent advancements have demonstrated the effectiveness of transformer models in framing decision-making as a sequence modeling task, especially noted for their scalability and efficiency within offline reinforcement learning contexts. Nonetheless, the extension of these models to the realm of Multi-Agent Reinforcement Learning (MARL) presents formidable obstacles, including credit assignment among individual agents while ensuring decentralized policy execution for practical deployment. This paper introduces Counterfactual Multi-agent Decison Transformer (CFMADT), a novel transformer-based approach for offline MARL within the Centralized Training with Decentralized Execution (CTDE) paradigm. CFMADT employs separate transformers for global and local trajectory modeling, using counterfactual reasoning to estimate individual agent contributions and facilitate effective credit assignment. The evaluation results on several complex offline MARL benchmarks, e.g., StarCraft Multi-Agent Challenge (SMAC) , SMAC v2 and Flatland, showcase that our method achieves exceptional performance.

An Agentic AI Workflow (AAW), also known as an LLM-based multi-agent system, is an autonomous system that assembles several LLM-based agents to work collaboratively towards a shared goal. The high autonomy, widespread adoption, and growing interest in such AAWs highlight the need for a deeper understanding of their operations, from both quality and security aspects. To this day, there are no existing methods to assess the influence of each agent on the AAW's final output. Adopting techniques from related fields is not feasible since existing methods perform only static structural analysis, which is unsuitable for inference time execution. We present Counterfactual-based Agent Influence Ranker (CAIR) - the first method for assessing the influence level of each agent on the AAW's output and determining which agents are the most influential. By performing counterfactual analysis, CAIR provides a task-agnostic analysis that can be used both offline and at inference time. We evaluate CAIR using an AAWs dataset of our creation, containing 30 different use cases with 230 different functionalities. Our evaluation showed that CAIR produces consistent rankings, outperforms baseline methods, and can easily enhance the effectiveness and relevancy of downstream tasks.

Causal inference holds immense value in fields such as healthcare, economics, and social sciences. However, traditional causal analysis workflows impose significant technical barriers, requiring researchers to possess dual backgrounds in statistics and computer science, while manually selecting algorithms, handling data quality issues, and interpreting complex results. To address these challenges, we propose CausalAgent, a conversational multi-agent system for end-to-end causal inference. The system innovatively integrates Multi-Agent Systems (MAS), Retrieval-Augmented Generation (RAG), and the Model Context Protocol (MCP) to achieve automation from data cleaning and causal structure learning to bias correction and report generation through natural language interaction. Users need only upload a dataset and pose questions in natural language to receive a rigorous, interactive analysis report. As a novel user-centered human-AI collaboration paradigm, CausalAgent explicitly models the analysis workflow. By leveraging interactive visualizations, it significantly lowers the barrier to entry for causal analysis while ensuring the rigor and interpretability of the process.

Abstract Understanding causality in medical research is essential for developing effective interventions and diagnostic tools. Mendelian Randomization (MR) is a pivotal method for inferring causality through genetic data. However, MR analysis often requires pre-identification of exposure-outcome pairs from clinical experience or literature, which can be challenging to obtain. This poses difficulties for clinicians investigating causal factors of specific diseases. To address this, we introduce MRAgent, an innovative automated agent leveraging Large Language Models (LLMs) to enhance causal knowledge discovery in disease research. MRAgent autonomously scans scientific literature, discovers potential exposure-outcome pairs, and performs MR causal inference using extensive Genome-Wide Association Study data. We conducted both automated and human evaluations to compare different LLMs in operating MRAgent and provided a proof-of-concept case to demonstrate the complete workflow. MRAgent’s capability to conduct large-scale causal analyses represents a significant advancement, equipping researchers and clinicians with a robust tool for exploring and validating causal relationships in complex diseases. Our code is public at https://github.com/xuwei1997/MRAgent.

This study examines the incorporation of causal inference methods into intelligent entities and examines the benefits of utilizing causal reasoning to improve decision-making procedures. This study entails conducting an experimental evaluation within a video game setting to evaluate the performance of three separate agent types: ExplorerBOT, GuardBOT, and CausalBOT. The ExplorerBOT utilizes a stochastic path selection technique for task completion, whereas the GuardBOT remains immobile yet exhibits exceptional proficiency in identifying and neutralizing other bots. On the other hand, the CausalBOT utilizes sophisticated causal inference methods to examine the underlying factors contributing to the failures noticed in the task completion of the ExplorerBOT. The aforementioned feature allows CausalBOT to make informed decisions by selecting paths that have a greater likelihood of achieving success. The main purpose of these experiments is to assess and compare the effectiveness of two distinct bots, namely ExplorerBOT and CausalBOT, in accomplishing their respective objectives. To facilitate comparison, two iterations of the ExplorerBOT are utilized. The initial iteration is predicated exclusively on stochastic path selection and necessitates a more profound understanding of the variables that impact the achievement of tasks. On the other hand, the second version integrates an algorithm for informed search. In contrast, CausalBOT employs causal inference techniques to discover the underlying causes of failures exhibited by ExplorerBOTs and collect pertinent data. Through the process of discerning the fundamental causal mechanisms, CausalBOT is able to make well-informed decisions by selecting pathways that maximize the probability of successfully completing a given job. The utilization of this approach greatly boosts the decision-making powers of CausalBOT, hence enabling it to effectively adapt and overcome problems in a more efficient manner when compared to alternative agents.

No abstract available

Causal analysis plays a foundational role in scientific discovery and reliable decision-making, yet it remains largely inaccessible to domain experts due to its conceptual and algorithmic complexity. This disconnect between causal methodology and practical usability presents a dual challenge: domain experts are unable to leverage recent advances in causal learning, while causal researchers lack broad, real-world deployment to test and refine their methods. To address this, we introduce Causal-Copilot, an autonomous agent that operationalizes expert-level causal analysis within a large language model framework. Causal-Copilot automates the full pipeline of causal analysis for both tabular and time-series data -- including causal discovery, causal inference, algorithm selection, hyperparameter optimization, result interpretation, and generation of actionable insights. It supports interactive refinement through natural language, lowering the barrier for non-specialists while preserving methodological rigor. By integrating over 20 state-of-the-art causal analysis techniques, our system fosters a virtuous cycle -- expanding access to advanced causal methods for domain experts while generating rich, real-world applications that inform and advance causal theory. Empirical evaluations demonstrate that Causal-Copilot achieves superior performance compared to existing baselines, offering a reliable, scalable, and extensible solution that bridges the gap between theoretical sophistication and real-world applicability in causal analysis. A live interactive demo of Causal-Copilot is available at https://causalcopilot.com/.

Value decomposition plays a pivotal role in ensuring effective credit assignment within Multi-Agent Reinforcement Learning (MARL), particularly in cooperative multi-agent tasks where agents are limited to accessing team rewards only. However, existing methods treat the mixing network as a black box, implicitly assuming that neural networks can autonomously extract important information and achieve rational credit assignment during policy learning. This approach not only lacks interpretability but may also prove inefficient in complex scenarios. To enhance the interpretability and rationality of value decomposition, we propose an innovative approach called “Causal Value Decomposition”(CMIX). CMIX employs causal inference-based models, introducing a set of metrics beyond environmental rewards to enhance robustness and model interpretability. Specifically, CMIX establishes intricate relational structures among agents in complex environments and leverages causal relationships between agents and their surroundings to address the credit assignment challenge in MARL. By employing do-calculus, CMIX accurately measures the impact of each agent's actions on environmental states, precisely determining their contribution to the collective reward. This approach not only enhances the interpretability of existing black-box models but also improves the accuracy of credit assignment in multi-agent systems. Moreover, CMIX exhibits high scalability and complements existing value decomposition techniques. Its effectiveness and scalability have been rigorously tested across various settings, including MPE, LBF, and SMAC environments.

In computed tomography (CT), achieving high image quality while minimizing radiation exposure remains a key clinical challenge. This paper presents CAPRI-CT, a novel causal-aware deep learning framework for Causal Analysis and Predictive Reasoning for Image Quality Optimization in CT imaging. CAPRI-CT integrates image data with acquisition metadata (such as tube voltage, tube current, and contrast agent types) to model the underlying causal relationships that influence image quality. An ensemble of Variational Autoencoders (VAEs) is employed to extract meaningful features and generate causal representations from observational data, including CT images and associated imaging parameters. These input features are fused to predict the Signal-to-Noise Ratio (SNR) and support counterfactual inference, enabling what-if simulations, such as changes in contrast agents (types and concentrations) or scan parameters. CAPRI-CT is trained and validated using an ensemble learning approach, achieving strong predictive performance. By facilitating both prediction and interpretability, CAPRI-CT provides actionable insights that could help radiologists and technicians design more efficient CT protocols without repeated physical scans. The source code and dataset are publicly available at https://github.com/SnehaGeorge22/capri-ct.

Recent text-to-video generation models have made remarkable progress in visual realism, motion fidelity, and text-video alignment, yet they still struggle to produce socially coherent behavior. Unlike humans, who readily infer intentions, beliefs, emotions, and social norms from brief visual cues, current models often generate literal scenes without capturing the underlying causal and psychological dynamics. To systematically assess this limitation, we introduce the first benchmark for social reasoning in video generation. Grounded in developmental and social psychology, the benchmark covers thirty classic social cognition paradigms spanning seven core dimensions: mental-state inference, goal-directed action, joint attention, social coordination, prosocial behavior, social norms, and multi-agent strategy. To operationalize these paradigms, we build a fully training-free agent-based pipeline that distills the reasoning structure of each paradigm, synthesizes diverse video-ready scenarios, enforces conceptual neutrality and difficulty control through cue-based critique, and evaluates generated videos with a high-capacity VLM judge along five interpretable dimensions of social reasoning. Using this framework, we conduct the first large-scale evaluation of seven state-of-the-art video generation systems. Results show a clear gap between surface-level plausibility and deeper social reasoning, suggesting that current models remain limited in their ability to generate socially grounded behavior. https://github.com/Gloria2tt/SVBench-Evaluation

We introduce a mathematically rigorous framework for an artificial intelligence system composed of probabilistic agents evolving through structured competition and belief revision. The architecture, grounded in Bayesian inference, measure theory, and population dynamics, defines agent fitness as a function of alignment with a fixed external oracle representing ground truth. Agents compete in a discrete-time environment, adjusting posterior beliefs through observed outcomes, with higher-rated agents reproducing and lower-rated agents undergoing extinction. Ratings are updated via pairwise truth-aligned utility comparisons, and belief updates preserve measurable consistency and stochastic convergence. We introduce hash-based cryptographic identity commitments to ensure traceability, alongside causal inference operators using do-calculus. Formal theorems on convergence, robustness, and evolutionary stability are provided. The system establishes truth as an evolutionary attractor, demonstrating that verifiable knowledge arises from adversarial epistemic pressure within a computable, self-regulating swarm.

The generalization capability of deepfake detectors is critical for real-world use. Data augmentation via synthetic fake face generation effectively enhances generalization, yet current SoTA methods rely on fixed strategies-raising a key question: Is a single static augmentation sufficient, or does the diversity of forgery features demand dynamic approaches? We argue existing methods overlook the evolving complexity of real-world forgeries (e.g., facial warping, expression manipulation), which fixed policies cannot fully simulate. To address this, we propose CRDA (Curriculum Reinforcement-Learning Data Augmentation), a novel framework guiding detectors to progressively master multi-domain forgery features from simple to complex. CRDA synthesizes augmented samples via a configurable pool of forgery operations and dynamically generates adversarial samples tailored to the detector's current learning state. Central to our approach is integrating reinforcement learning (RL) and causal inference. An RL agent dynamically selects augmentation actions based on detector performance to efficiently explore the vast augmentation space, adapting to increasingly challenging forgeries. Simultaneously, the agent introduces action space variations to generate heterogeneous forgery patterns, guided by causal inference to mitigate spurious correlations-suppressing task-irrelevant biases and focusing on causally invariant features. This integration ensures robust generalization by decoupling synthetic augmentation patterns from the model's learned representations. Extensive experiments show our method significantly improves detector generalizability, outperforming SOTA methods across multiple cross-domain datasets.

Modern manufacturing systems demand real-time, trustworthy, and interpretable insights into anomalies and their underlying causes. However, conventional pipelines treat anomaly detection, causal inference, and decision-making as siloed tasks, lacking integration, explainability, and adaptability. We present CausalPulse, an intelligent, multi-agent copilot for automated Root Cause Analysis (RCA) in industrial settings. Built on a modular and extensible architecture, the system leverages standard agentic protocols, including Model Context Protocol (MCP), Agent2Agent (A2A), and LangGraph for dynamic tool and agent discovery and seamless orchestration of tasks. Agents dynamically interact to perform data preprocessing, anomaly detection, causal discovery, and root cause analysis through a neurosymbolic workflow that combines symbolic reasoning with neural methods. Intelligent postprocessing pipelines enable automatic chaining of agent tasks, enhancing contextual awareness and adaptability. CausalPulse is evaluated using both an academic public dataset (i.e., Future Factories) and an industrial proprietary dataset (i.e., Planar Oxygen Sensor Element) and shows that the system outperforms traditional baselines in interpretability, trustworthiness, and operational utility.

Despite advancements in causal inference and prescriptive AI, its adoption in enterprise settings remains hindered primarily due to its technical complexity. Many users lack the necessary knowledge and appropriate tools to effectively leverage these technologies. This work at the MIT-IBM Watson AI Lab focuses on developing the proof-of-concept agent, PrecAIse, a domain-adaptable conversational agent equipped with a suite of causal and prescriptive tools to help enterprise users make better business decisions. The objective is to make advanced, novel causal inference and prescriptive tools widely accessible through natural language interactions. The presented Natural Language User Interface (NLUI) enables users with limited expertise in machine learning and data science to harness prescriptive analytics in their decision-making processes without requiring intensive computing resources. We present an agent capable of function calling, maintaining faithful, interactive, and dynamic conversations, and supporting new domains.

Recent applications of artificial intelligence in economics and finance have been dominated by predictive machine learning approaches that deliver impressive forecasting performance in stable environments but offer limited support for explanation, policy evaluation, and structural change analysis. While such models excel at detecting correlations, they struggle to address causal mechanisms, counterfactual reasoning, and emergent dynamics that are central to economic decision-making and policy design. This paper argues that these limitations stem from an overreliance on prediction-oriented AI and proposes an integrated framework that combines causal AI, generative AI, and agent-based AI to better align artificial intelligence with economic reasoning. The framework synthesizes advances in causal inference, deep generative modeling, and computational economics to move beyond black-box prediction toward systems capable of explanation, simulation, and intervention. Specifically, the proposed approach emphasizes three capabilities that purely predictive models lack: identification of policy-invariant causal relationships, robust counterfactual and stress-test analysis under structural change, and the modeling of emergent macroeconomic outcomes arising from heterogeneous agent interactions. By unifying these paradigms, the paper provides a conceptual foundation for AI systems that support policy evaluation, scenario analysis, and institutional design in complex economic and financial systems. The paper concludes by outlining ethical, governance, and institutional considerations and by proposing a research agenda for developing economically grounded, transparent, and policy-relevant AI tools that complement rather than replace economic theory.

Failures are the norm in highly complex and heterogeneous devices spanning the distributed computing continuum (DCC), from resource-constrained IoT and edge nodes to high-performance computing systems. Ensuring reliability and global consistency across these layers remains a major challenge, especially for AI-driven workloads requiring real-time, adaptive coordination. This work-in-progress paper introduces a Probabilistic Active Inference Resilience Agent (PAIR-Agent) to achieve resilience in DCC systems. PAIR-Agent performs three core operations: (i) constructing a causal fault graph from device logs, (ii) identifying faults while managing certainties and uncertainties using Markov blankets and the free energy principle, and (iii) autonomously healing issues through active inference. Through continuous monitoring and adaptive reconfiguration, the agent maintains service continuity and stability under diverse failure conditions. Theoretical validations confirm the reliability and effectiveness of the proposed framework.

Large language models (LLMs), while capable of generating plausible diagnostic plans from sensor data, inherently lack true causal reasoning capabilities and are prone to hallucinations. To address this limitation, we propose a hybrid AI framework that integrates LLMs with structured causal inference to enable robust, interpretable decision-making in predictive health monitoring (PHM) for complex systems. Our architecture positions the LLM as a planning agent that infers candidate failure modes and troubleshooting steps, while delegating causal evaluation to an external inference model grounded in formal causal principles. The system constructs a localized causal knowledge graph (KG) by retrieving top-k similar historical traces based on the current sensor context, and uses this graph to simulate and evaluate the impact of potential actions. We explore three strategies for handling multi-step diagnostic plans: step-wise decomposition, compound treatment modelling, and sequential intervention chains. Recommendations are ranked based on their estimated effect on resolution likelihood and further validated by a dedicated Evaluator Agent using counterfactual reasoning. Our results demonstrate that augmenting LLM-generated plans with external causal inference significantly improves relevance, consistency, and safety—offering a deployable blueprint for high-stakes PHM scenarios where LLMs alone cannot be trusted to reason reliably.

Artificial Intelligence has achieved remarkable advancements in recent years, yet much of its progress relies on identifying increasingly complex correlations. Enabling causality awareness in AI has the potential to enhance its performance by enabling a deeper understanding of the underlying mechanisms of the environment. In this paper, we introduce DODO, an algorithm defining how an Agent can autonomously learn the causal structure of its environment through repeated interventions. We assume a scenario where an Agent interacts with a world governed by a causal Directed Acyclic Graph (DAG), which dictates the system's dynamics but remains hidden from the Agent. The Agent's task is to accurately infer the causal DAG, even in the presence of noise. To achieve this, the Agent performs interventions, leveraging causal inference techniques to analyze the statistical significance of observed changes. Results show better performance for DODO, compared to observational approaches, in all but the most limited resource conditions. DODO is often able to reconstruct with as low as zero errors the structure of the causal graph. In the most challenging configuration, DODO outperforms the best baseline by +0.25 F1 points.

Multiscale interactions between genes, cells, communities, hosts, and environments result in Antimicrobial Resistance (AMR), which are hard to predict using classical, single-paradigm models. Our concept of a combined predictive microbiological framework is a synthesis of mechanistic kinetics and Pharmacodynamics/Pharmacokinetics (PK/PK) limits and the present-day AI to predict microbial growth and death, as well as resistance development. The biophysical structure and safety limits are put in place through the mechanistic core ordinary differential equations, reaction-diffusion transport, and agent-based biofilm modules. Graph Neural Networks(GNNs) data-driven Components graph neural networks over gene-drug-plasmid graphs, Protein/DNA language models, Resistome profiling Bayesian deep learners, Minimum Inhibitory Concentration (MIC) regression provide flexible function approximation with calibrated uncertainty. These layers are connected by probabilistic data assimilation, and the dosing strategies are assessed, and causal inference and counterfactual simulation attribute resistance mechanisms (efflux, target modification, enzymatic degradation, and permeability changes). Active learning will pick experiments (e.g., time-kill assays, lab-on-chip gradients) that minimize posterior uncertainty to the maximum, thus leading to an iterative digital twin of microbiology. Validate across stratified pathogen-site splits and external challenge sets, demonstrating improved MIC accuracy, better calibration, and more faithful phase timing versus sequence-only or kinetics-only baselines. The framework assists with stewardship and food-safety decisions by predicting the outcome of treatments using mono- and combination therapies, the collateral sensitivity, and stress-testing the policies in conditions of environmental variability. This method is based on interpretable biophysics and scalable AI to speed up hypothesis generation, optimize antibiotic regimens, and enhance surveillance in AMR. Predictive Microbiology, Antimicrobial Resistance (AMR), MIC Prediction, Agent-Based Biofilm Simulation, Graph Neural Networks, Active Learning.

The culmination of explainable AI, causal inference, and reinforcement learning creates a radical departure from convention in creating a systematic approach to precisely allocate and manage multi-channel marketing budgets in real time. Indeed, it marks the culmination of increasingly complex calls and demands from modern marketing environments. The framework optimizes the data pipelines of ad platforms, CRM, promotions, and external signals, welded together with ensemble machine-learning models and Bayesian calibration for optimally accurate marginal ROI and uplift forecasts. Causal modeling, formed out of geo- and switchback experiments, serves to isolate incremental effects while minimizing attribution bias arising from legacy systems. Budget optimization and allocation actions flow from a constrained optimization layer that translates those insights, dynamically enlivened through an online RL agent learning from feedback-on-realized returns. The system combines SHAP interpretability with interactive dashboards to ensure transparency and trust by the marketing community. A lot of empirical model testing on the data sets from the industry was done for high predictive accuracy, incremental ROI estimation, budget optimization, and differential satisfaction compared to the traditional benchmark. Whereas this is cloud-based and modularly secure for deployment, the architecture should be scalable, privacy-compliant, and ready. This is a step forward toward evidence-based operational decision support in marketing analytics that targets optimally efficient, ethical, and transparentspending in a quickly shifting environment.

This paper introduces the Impact-Driven AI Framework (IDAIF), a novel architectural methodology that integrates Theory of Change (ToC) principles with modern artificial intelligence system design. As AI systems increasingly influence high-stakes domains including healthcare, finance, and public policy, the alignment problem--ensuring AI behavior corresponds with human values and intentions--has become critical. Current approaches predominantly optimize technical performance metrics while neglecting the sociotechnical dimensions of AI deployment. IDAIF addresses this gap by establishing a systematic mapping between ToC's five-stage model (Inputs-Activities-Outputs-Outcomes-Impact) and corresponding AI architectural layers (Data Layer-Pipeline Layer-Inference Layer-Agentic Layer-Normative Layer). Each layer incorporates rigorous theoretical foundations: multi-objective Pareto optimization for value alignment, hierarchical multi-agent orchestration for outcome achievement, causal directed acyclic graphs (DAGs) for hallucination mitigation, and adversarial debiasing with Reinforcement Learning from Human Feedback (RLHF) for fairness assurance. We provide formal mathematical formulations for each component and introduce an Assurance Layer that manages assumption failures through guardian architectures. Three case studies demonstrate IDAIF application across healthcare, cybersecurity, and software engineering domains. This framework represents a paradigm shift from model-centric to impact-centric AI development, providing engineers with concrete architectural patterns for building ethical, trustworthy, and socially beneficial AI systems.

AI and population health are becoming increasingly intertwined, driven by the growing availability of multimodal data and rapid advances in AI. At the AAAI-26 New Faculty Highlights, I present our efforts to harness these trends to enhance our capacity to model, simulate, and adapt to complex dynamical processes. I first introduce our robust deep learning architectures for real-time outbreak response, highlighting how our frameworks capture uncertainty and dynamics across shifting distributions, multimodal data, hierarchical structures, and relational dependencies. I will then introduce our hybrid approaches that integrate machine learning with science-based mechanistic epidemiological models, including physics-informed neural networks, expert-guided generative models for causal inference, and differentiable agent-based models. Together, these advances illustrate how combining data-driven AI with domain knowledge can enable more reliable, adaptive, and actionable solutions to inform decision making in population health.

Modern automotive manufacturing demands minimal downtime and near-zero defects. This paper proposes an agentic AI framework for self-healing production lines, enabling autonomous fault diagnosis and correction in real time. We integrate agent-based models, causal inference, and self-adaptive control to monitor processes, identify root causes of faults, and adjust controls without human intervention. The architecture is demonstrated in scenarios like body-in-white welding and final assembly inspection, where AI agents at robotic stations collaborate with a supervisor agent to detect quality issues (e.g., bad welds, misalignments), infer underlying causes, and enact corrective actions (like recalibration or parameter tuning). A novel case study is presented in which an autonomous welding cell agent detects a weld defect, determines tip wear as root cause, and triggers an on-the-fly tool change and re-weld—preventing downtime. We report substantial improvements: fault response times drop from tens of minutes to seconds, process recovery becomes nearly instantaneous, and overall equipment effectiveness (OEE) rises with reduced scrap and downtime. Five high-quality images, three charts, and three diagrams illustrate the agentic system architecture, decision loops, fault response performance, and comparative benchmarks. The proposed framework—unlike any published to date—demonstrates a unique, self-healing manufacturing AI that achieves resilient, “right-first-time” production in automotive assembly.

Failure attribution in multi-agent systems -- pinpointing the exact step where a decisive error occurs -- is a critical yet unsolved challenge. Current methods treat this as a pattern recognition task over long conversation logs, leading to critically low step-level accuracy (below 17\%), which renders them impractical for debugging complex systems. Their core weakness is a fundamental inability to perform robust counterfactual reasoning: to determine if correcting a single action would have actually averted the task failure. To bridge this \emph{counterfactual inference gap}, we introduce Abduct-Act-Predict (A2P) Scaffolding, a novel agent framework that transforms failure attribution from pattern recognition into a structured causal inference task. A2P explicitly guides a large language model through a formal three-step reasoning process within a single inference pass: (1) Abduction, to infer the hidden root causes behind an agent's actions; (2) Action, to define a minimal corrective intervention; and (3) Prediction, to simulate the subsequent trajectory and verify if the intervention resolves the failure. This structured approach leverages the holistic context of the entire conversation while imposing a rigorous causal logic on the model's analysis. Our extensive experiments on the Who\&When benchmark demonstrate its efficacy. On the Algorithm-Generated dataset, A2P achieves 47.46\% step-level accuracy, a 2.85$\times$ improvement over the 16.67\% of the baseline. On the more complex Hand-Crafted dataset, it achieves 29.31\% step accuracy, a 2.43$\times$ improvement over the baseline's 12.07\%. By reframing the problem through a causal lens, A2P Scaffolding provides a robust, verifiable, and significantly more accurate solution for automated failure attribution. Ours code are released at https://github.com/ResearAI/A2P.

Multi-agent systems (MAS) are critical for automating complex tasks, yet their practical deployment is severely hampered by the challenge of failure attribution. Current diagnostic tools, which rely on statistical correlations, are fundamentally inadequate; on challenging benchmarks like Who\&When, state-of-the-art methods achieve less than 15\% accuracy in locating the root-cause step of a failure. To address this critical gap, we introduce the first failure attribution framework for MAS grounded in multi-granularity causal inference. Our approach makes two key technical contributions: (1) a performance causal inversion principle, which correctly models performance dependencies by reversing the data flow in execution logs, combined with Shapley values to accurately assign agent-level blame; (2) a novel causal discovery algorithm, CDC-MAS, that robustly identifies critical failure steps by tackling the non-stationary nature of MAS interaction data. The framework's attribution results directly fuel an automated optimization loop, generating targeted suggestions whose efficacy is validated via counterfactual simulations. Evaluations on the Who\&When and TRAIL benchmarks demonstrate a significant leap in performance. Our method achieves up to 36.2\% step-level accuracy. Crucially, the generated optimizations boost overall task success rates by an average of 22.4\%. This work provides a principled and effective solution for debugging complex agent interactions, paving the way for more reliable and interpretable multi-agent systems.

Multi-Agent Reinforcement Learning (MARL) has demonstrated significant potential in the field of collaborative decision-making. However, its performance is often limited by environmental non-stationarity and complex interdependencies among agents. Traditional methods typically learn policies based on correlations, which can lead to the acquisition of spurious associations, resulting in poor policy generalization and insufficient robustness. To address these issues, this paper proposes a novel Multi-Agent Policy Optimization algorithm integrating Causal Inference (Causal-MAPO). The core of this algorithm lies in constructing a differentiable causal graph model to explicitly model the causal structure between agent actions and joint rewards. By introducing an attention-based mechanism to infer the causal graph from observational data and utilizing this graph to guide credit assignment and information sharing among agents during training, the algorithm filters out non-causal interference signals. Experimental results on the StarCraft II Multi-Agent Challenge (SMAC) and a simulated traffic network environment demonstrate that compared to baseline algorithms, Causal-MAPO achieves higher win rates and sample efficiency in complex tasks. Furthermore, it exhibits significantly enhanced decision-making robustness and generalization capability when facing dynamic disturbances and previously unseen opponent strategies.

Under the “Dual Carbon” goal, public opinion analysis is crucial for optimizing policy implementation and enhancing social consensus, yet it faces challenges such as insufficient multi-source data integration, limited causal modeling, and delayed interventions. This study proposes a collaborative framework integrating reinforcement learning-enhanced large language models (LLMs), diffusion models, and multi-agent systems (MASs). By constructing a four-dimensional causal network of “policy–technology–economy–public sentiment”, it analyzes multi-source data and simulates multi-agent interactions. The experimental results show that this framework outperforms Latent Dirichlet Allocation (LDA), Bidirectional Encoder Representations from Transformers (BERT), and Susceptible Infected Recovered (SIR) models in causal inference, dynamic intervention, and multi-agent collaboration. Reinforcement Learning from Human Feedback (RLHF) optimizes LLM outputs for reliable policy recommendations, with pass@10 showing strong correlations. This study provides scientific support for “Dual Carbon” policymaking and public opinion guidance, facilitating the green and low-carbon transition.

We address the challenge of explaining counterfactual outcomes in multi-agent Markov decision processes. In particular, we aim to explain the total counterfactual effect of an agent's action on the outcome of a realized scenario through its influence on the environment dynamics and the agents'behavior. To achieve this, we introduce a novel causal explanation formula that decomposes the counterfactual effect by attributing to each agent and state variable a score reflecting their respective contributions to the effect. First, we show that the total counterfactual effect of an agent's action can be decomposed into two components: one measuring the effect that propagates through all subsequent agents'actions and another related to the effect that propagates through the state transitions. Building on recent advancements in causal contribution analysis, we further decompose these two effects as follows. For the former, we consider agent-specific effects -- a causal concept that quantifies the counterfactual effect of an agent's action that propagates through a subset of agents. Based on this notion, we use Shapley value to attribute the effect to individual agents. For the latter, we consider the concept of structure-preserving interventions and attribute the effect to state variables based on their"intrinsic"contributions. Through extensive experimentation, we demonstrate the interpretability of our approach in a Gridworld environment with LLM-assisted agents and a sepsis management simulator.

Decision Transformers (DT) play a crucial role in modern reinforcement learning, leveraging offline datasets to achieve impressive results across various domains. However, DT requires high-quality, comprehensive data to perform optimally. In real-world applications, the lack of training data and the scarcity of optimal behaviours make training on offline datasets challenging, as suboptimal data can hinder performance. To address this, we propose the Counterfactual Reasoning Decision Transformer (CRDT), a novel framework inspired by counterfactual reasoning. CRDT enhances DT’s ability to reason beyond known data by generating and utilizing counterfactual experiences, enabling improved decision-making in unseen scenarios. Experiments across Atari and D4RL benchmarks, including scenarios with limited data and altered dynamics, demonstrate that CRDT outperforms conventional DT approaches. Additionally, reasoning counterfactually allows the DT agent to obtain stitching abilities, combining suboptimal trajectories, without architectural modifications. These results highlight the potential of counterfactual reasoning to enhance reinforcement learning agents' performance and generalization capabilities.

In modern communication, the electromagnetic spectrum serves as the carrier for information transmission, and the only medium enabling information exchange anywhere, anytime. To adapt to the changing dynamics of a complex electromagnetic environment, electromagnetic spectrum allocation algorithms must not only meet the demands for efficiency and intelligence but also possess anti-jamming capabilities to achieve the best communication effect. Focusing on intelligent wireless communication, this paper proposes a multi-agent hybrid game spectrum allocation method under incomplete information and based on the Monte Carlo counter-factual regret minimization algorithm. Specifically, the method first utilizes frequency usage and interference information from both sides to train agents through extensive simulations using the Monte Carlo Method, allowing the trial values to approach the expected values. Based on the results of each trial, the counterfactual regret minimization algorithm is employed to update the frequency selection strategies for both the user and the interferer. Subsequently, the trained agents from both sides engage in countermeasure communication. Finally, the probabilities of successful communication and successful interference for both sides are statistically analyzed. The results show that under the multi-agent hybrid game spectrum allocation method based on the Monte Carlo counter-factual regret minimization algorithm, the probability of successful interference against the user is 32.5%, while the probability of successful interference by the jammer is 37.3%. The average simulation time per round is 3.06 s. This simulation validates the feasibility and effectiveness of the multi-agent hybrid game spectrum allocation module based on the Monte Carlo counter-factual regret minimization algorithm.

In the field of autonomous driving, safe and efficient decision-making through deep reinforcement learning remains a significant challenge. Existing methods often struggle to adapt to the dynamic and complex conditions of urban environments, while the lack of interpretability in reinforcement learning algorithms raises safety concerns. To address these issues, this paper proposes a reinforcement learning method guided by reward machine. Firstly, a reward machine tailored for autonomous driving scenarios is constructed to better guide the agent's behavior selection. Then, a time sensitivity factor is proposed to adjust the generation of counterfactual experiences, optimizing the effectiveness of policy learning. Furthermore, to improve the safety of autonomous driving decisions, formal verification of the decision-making process is conducted via the UPPAAL model checker, enabling the identification and handling of potential hazardous states. Finally, the effectiveness and safety of this method are validated through a case study of an autonomous driving system, demonstrating that the proposed reward machine-guided reinforcement learning algorithm performs well in complex road scenarios.

Agent-based simulation is crucial for modeling complex human behavior, yet traditional approaches require extensive domain knowledge and large datasets. In data-scarce healthcare settings where historic and counterfactual data are limited, large language models (LLMs) offer a promising alternative by leveraging broad world knowledge. This study examines an LLM-driven simulation of a maternal mobile health program, predicting beneficiaries' listening behavior when they receive health information via automated messages (control) or live representatives (intervention). Since uncertainty quantification is critical for decision-making in health interventions, we propose an LLM epistemic uncertainty estimation method based on binary entropy across multiple samples. We enhance model robustness through ensemble approaches, improving F1 score and model calibration compared to individual models. Beyond direct evaluation, we take a decision-focused approach, demonstrating how LLM predictions inform intervention feasibility and trial implementation in data-limited settings. The proposed method extends to public health, disaster response, and other domains requiring rapid intervention assessment under severe data constraints. All code and prompts used for this work can be found at https://github.com/sarahmart/LLM-ABS-ARMMAN-prediction.

In complex dynamic environment, there are some problems in multi-agent collaborative decision-making, such as non-stationarity, credit distribution deviation and imperfect topological relationship modeling. In this article, the deep reinforcement learning (DRL) algorithm Att-GNN-MARL, which combines attention mechanism and GNN, is proposed to improve the cooperative efficiency and strategic stability of multi-agent system (MAS). Based on the centralized training decentralized execution (CTDE) framework, this study uses graph neural network (GNN) to model the spatial connection of agents, introduces attention mechanism to enhance the perception of key neighbors, and designs a counterfactual reward correction mechanism to optimize local credit allocation. In the collaborative control scenario of urban traffic signals, the experiments based on real road network on SUMO simulation platform show that the proposed algorithm reduces the average waiting time of vehicles to 54.3 seconds during rush hours, which is 22.3% less than that of QMIX. The total traffic volume reached 24,108 vehicles, which was significantly better than the comparison method. The number of rounds of strategy convergence is 640, and the learning efficiency is good. The research shows that the cooperative learning mechanism combining structural perception and dynamic attention can improve the overall performance of MAS in highly coupled tasks to some extent.

Given the rapid advancement of artificial intelligence, understanding the foundations of intelligent behaviour is increasingly important. Active inference, regarded as a general theory of behaviour, offers a principled approach to probing the basis of sophistication in planning and decision-making. This paper examines two decision-making schemes in active inference based on “planning” and “learning from experience”. Furthermore, we also introduce a mixed model that navigates the data complexity trade-off between these strategies, leveraging the strengths of both to facilitate balanced decision-making. We evaluate our proposed model in a challenging grid-world scenario that requires adaptability from the agent. Additionally, our model provides the opportunity to analyse the evolution of various parameters, offering valuable insights and contributing to an explainable framework for intelligent decision-making.

Introduction To address the challenges of data heterogeneity, strategic diversity, and process opacity in interpreting multi-agent decision-making within complex competitive environments, we have developed TRACE, an end-to-end analytical framework for StarCraft II gameplay. Methods This framework standardizes raw replay data into aligned state trajectories, extracts “typical strategic progressions” using a Conditional Recurrent Variational Autoencoder (C-RVAE), and quantifies the deviation of individual games from these archetypes via counterfactual alignment. Its core innovation is the introduction of a dimensionless deviation metric, |Δ|, which achieves process-level interpretability. This metric reveals “which elements are important” by ranking time-averaged feature contributions across aggregated categories (Economy, Military, Technology) and shows “when deviations occur” through temporal heatmaps, forging a verifiable evidence chain.. Results Quantitative evaluation on professional tournament datasets demonstrates the framework’s robustness, revealing that strategic deviations often crystallize in the early game (averaging 8.4% of match duration) and are frequently driven by critical technology timing gaps. The counterfactual generation module effectively restores strategic alignment, achieving an average similarity improvement of over 90% by correcting identified divergences. Furthermore, expert human evaluation confirms the practical utility of the system, awarding high scores for Factual Fidelity (4.6/5.0) and Causal Coherence (4.3/5.0) to the automatically generated narratives. Discussion By providing openaccess code and reproducible datasets, TRACE lowers the barrier to large-scale replay analysis, offering an operational quantitative basis for macro-strategy understanding, coaching reviews, and AI model evaluation.

In this work, we consider AI agents operating in Partially Observable Markov Decision Processes (POMDPs)-a widely-used framework for sequential decision making with incomplete state information. Agents operating with partial information take actions not only to advance their underlying goals but also to seek information and reduce uncertainty. Despite rapid progress in explainable AI, research on separating information-driven vs. goal-driven behaviors remains sparse. To address this gap, we introduce a novel explanation generation framework called Sequential Information Probing (SIP), to investigate the direct impact of state information, or its absence, on agent behavior. To quantify the impact we also propose two metrics under this SIP framework called Value of Information (VoI) and Influence of Information (IoI). We then theoretically derive several properties of these metrics. Finally, we present several experiments, including a case study on an autonomous vehicle, that illustrate the efficacy of our method.

We introduce the General Incentives-based Framework for Fairness (GIFF), a novel approach for fair multi-agent resource allocation that infers fair decision-making from standard value functions. In resource-constrained settings, agents optimizing for efficiency often create inequitable outcomes. Our approach leverages the action-value (Q-)function to balance efficiency and fairness without requiring additional training. Specifically, our method computes a local fairness gain for each action and introduces a counterfactual advantage correction term to discourage over-allocation to already well-off agents. This approach is formalized within a centralized control setting, where an arbitrator uses the GIFF-modified Q-values to solve an allocation problem. Empirical evaluations across diverse domains, including dynamic ridesharing, homelessness prevention, and a complex job allocation task-demonstrate that our framework consistently outperforms strong baselines and can discover far-sighted, equitable policies. The framework's effectiveness is supported by a theoretical foundation; we prove its fairness surrogate is a principled lower bound on the true fairness improvement and that its trade-off parameter offers monotonic tuning. Our findings establish GIFF as a robust and principled framework for leveraging standard reinforcement learning components to achieve more equitable outcomes in complex multi-agent systems.

Multi-Agent Reinforcement Learning (MARL) has proven to be a compelling tool for addressing decision-making tasks with multiple agents who are engaged in complex interactions and collaborations. This study evaluates a variety of policy gradient techniques that are currently used in MARL, specifying three up-to-date methods: Counterfactual Multi-Agent Policy Gradient (COMA), Meta-Learning Policy Gradient (Meta-PG), and Status Quo Policy Gradient (SQPG). The performance of each method is determined by its convergence speed across training, success in varying environments, and the stability of the variance. The experimental results indicate that Meta-PG is the fastest algorithm, achieving the highest performance metrics in both shared and teamwork-based tasks, being optimal in such cases. COMA, however, exhibits great stability and effectiveness in adversarial settings; and employs the novel idea of counterfactual credit assignment for better learning. Though SQPG offers an overall balanced performance in every environment, it suffers from generally low variance and long convergence due to its equilibrium-seeking characteristic. These research results exhibit the trade-offs in terms of learning speed, stability, and adaptability in MARL. From the study, Meta-PG is noted to be effective for fast learning, COMA for adversarial interactions, while SQPG is a general method that needs more refinement. MARL's real-world applicability should be enhanced by focusing on hybrid models that combine the advantages of these approaches, improved variance reduction techniques, and iterative testing on a larger population of agents.

In the field of multi-agent reinforcement learning, optimizing global rewards while balancing individual agent decisions remains a key challenge. This paper proposes a novel multi-agent reinforcement learning algorithm, CPPO (Counterfactual Proximal Policy Optimization), for global reward optimization in Autonomous Mobility-on-Demand (AMoD) systems. By integrating the stability of Proximal Policy Optimization (PPO) with the credit assignment mechanism of Counterfactual Multi-Agent (COMA), the CPPO algorithm enhances collaborative decision-making through a counterfactual advantage evaluation mechanism. We validate our approach using real-world data from New York City taxi trips across various spatial configurations and fleet sizes. Experimental results demonstrate that compared to traditional greedy strategies and random assignment approaches, CPPO shows significant advantages in terms of average rewards and request rejection rates.

Real-world multi-agent systems are often dynamic and continuous, where the agents co-evolve and undergo changes in their trajectories and interactions over time. For example, the COVID-19 transmission in the U.S. can be viewed as a multi-agent system, where states act as agents and daily population movements between them are interactions. Estimating the counterfactual outcomes in such systems enables accurate future predictions and effective decision-making, such as formulating COVID-19 policies. However, existing methods fail to model the continuous dynamic effects of treatments on the outcome, especially when multiple treatments (e.g., "stay-at-home" and "get-vaccine" policies) are applied simultaneously. To tackle this challenge, we propose Causal Graph Ordinary Differential Equations (CAG-ODE), a novel model that captures the continuous interaction among agents using a Graph Neural Network (GNN) as the ODE function. The key innovation of our model is to learn time-dependent representations of treatments and incorporate them into the ODE function, enabling precise predictions of potential outcomes. To mitigate confounding bias, we further propose two domain adversarial learning-based objectives, which enable our model to learn balanced continuous representations that are not affected by treatments or interference. Experiments on two datasets (i.e., COVID-19 and tumor growth) demonstrate the superior performance of our proposed model.

Fairness in Multi-Agent Systems (MAS) has been extensively studied, particularly in reward distribution among agents in scenarios such as goods allocation, resource division, lotteries, and bargaining systems. Fairness in MAS depends on various factors, including the system's governing rules, the behaviour of the agents, and their characteristics. Yet, fairness in human society often involves evaluating disparities between disadvantaged and privileged groups, guided by principles of Equality, Diversity, and Inclusion (EDI). Taking inspiration from the work on algorithmic fairness, which addresses bias in machine learning-based decision-making, we define protected attributes for MAS as characteristics that should not disadvantage an agent in terms of its expected rewards. We adapt fairness metrics from the algorithmic fairness literature -- namely, demographic parity, counterfactual fairness, and conditional statistical parity -- to the multi-agent setting, where self-interested agents interact within an environment. These metrics allow us to evaluate the fairness of MAS, with the ultimate aim of designing MAS that do not disadvantage agents based on protected attributes.

Multi-agent cyber-physical systems are present in a variety of applications. Agent decision-making can be affected due to errors induced by uncertain, dynamic operating environments or due to incorrect actions taken by an agent. When an erroneous decision that leads to a violation of safety is identified, assigning responsibility to individual agents is a key step towards preventing future accidents. Current approaches to carrying out such investigations require human labor or high degree of familiarity with operating environments. Automated strategies to assign responsibility can achieve significant reduction in human effort and associated cognitive burden.In this paper, we develop an automated procedure to assign responsibility for safety violations to actions of any single agent in a principled manner. We ground our approach on reasoning about safety violations in road safety. When provided with an instance of a safety violation, we use counterfactual reasoning to create alternate scenarios that determine how different outcomes might have been achieved if a specific action or set of actions was replaced by another action or set of actions. We devise a metric called the degree of responsibility (DoR) for each agent. The DoR uses the Shapley value to quantify the relative contribution of each agent to the observed safety violation, thus serving as a basis to explain and justify future decisions. We devise both heuristic techniques and methods based on the structure of agent interactions to improve scalability of our solution as the number of agents increases. We consider three instances of safety violations from the National Highway Traffic Safety Administration (NHTSA). We carry out experiments using representations of the three scenarios using the CARLA urban driving simulator. Our results indicate that the DoR enhances explainability of decision-making and assigning accountability for actions of agents and their consequences.

Recent breakthroughs in Large Language Models (LLMs) have led to a qualitative leap in artificial intelligence' s performance on reasoning tasks, particularly demonstrating remarkable capabilities in mathematical, symbolic, and commonsense reasoning. However, as a critical component of advanced human cognition, strategic reasoning, i.e., the ability to assess multi-agent behaviors in dynamic environments, formulate action plans, and adapt strategies, has yet to be systematically evaluated or modeled. To address this gap, this paper introduces WGSR-Bench, the first strategy reasoning benchmark for LLMs using wargame as its evaluation environment. Wargame, a quintessential high-complexity strategic scenario, integrates environmental uncertainty, adversarial dynamics, and non-unique strategic choices, making it an effective testbed for assessing LLMs' capabilities in multi-agent decision-making, intent inference, and counterfactual reasoning. WGSR-Bench designs test samples around three core tasks, i.e., Environmental situation awareness, Opponent risk modeling and Policy generation, which serve as the core S-POE architecture, to systematically assess main abilities of strategic reasoning. Finally, an LLM-based wargame agent is designed to integrate these parts for a comprehensive strategy reasoning assessment. With WGSR-Bench, we hope to assess the strengths and limitations of state-of-the-art LLMs in game-theoretic strategic reasoning and to advance research in large model-driven strategic intelligence.

Fairness in multi-agent systems (MAS) focuses on equitable reward distribution among agents in scenarios involving sensitive attributes such as race, gender, or socioeconomic status. This paper introduces fairness in Proximal Policy Optimization (PPO) with a penalty term derived from a fairness definition such as demographic parity, counterfactual fairness, or conditional statistical parity. The proposed method, which we call Fair-PPO, balances reward maximisation with fairness by integrating two penalty components: a retrospective component that minimises disparities in past outcomes and a prospective component that ensures fairness in future decision-making. We evaluate our approach in two games: the Allelopathic Harvest, a cooperative and competitive MAS focused on resource collection, where some agents possess a sensitive attribute, and HospitalSim, a hospital simulation, in which agents coordinate the operations of hospital patients with different mobility and priority needs. Experiments show that Fair-PPO achieves fairer policies than PPO across the fairness metrics and, through the retrospective and prospective penalty components, reveals a wide spectrum of strategies to improve fairness; at the same time, its performance pairs with that of state-of-the-art fair reinforcement-learning algorithms. Fairness comes at the cost of reduced efficiency, but does not compromise equality among the overall population (Gini index). These findings underscore the potential of Fair-PPO to address fairness challenges in MAS.

Explainable AI (XAI) systems have been proposed to help people understand how AI systems produce outputs and behaviors. Explainable Reinforcement Learning (XRL) has an added complexity due to the temporal nature of sequential decision-making. Further, non-AI experts do not necessarily have the ability to alter an agent or its policy. We introduce a technique for using World Models to generate explanations for Model-Based Deep RL agents. World Models predict how the world will change when actions are performed, allowing for the generation of counterfactual trajectories. However, identifying what a user wanted the agent to do is not enough to understand why the agent did something else. We augment Model-Based RL agents with a Reverse World Model, which predicts what the state of the world should have been for the agent to prefer a given counterfactual action. We show that explanations that show users what the world should have been like significantly increase their understanding of the agent policy. We hypothesize that our explanations can help users learn how to control the agents execution through by manipulating the environment.

Differentiable simulators represent an environment’s dynamics as a differentiable function. Within robotics and autonomous driving, this property is used in Analytic Policy Gradients (APG), which relies on backpropagating through the dynamics to train accurate policies for diverse tasks. Here we show that differentiable simulation also has an important role in world modeling, where it can impart predictive, prescriptive, and counterfactual capabilities to an agent. Specifically, we design three novel task setups in which the differentiable dynamics are combined within an end-to-end computation graph not with a policy, but a state predictor. This allows us to learn relative odometry, optimal planners, and optimal inverse states. We collectively call these predictors Analytic World Models (AWMs) and demonstrate how differentiable simulation enables their efficient, end-to-end learning. In autonomous driving scenarios, they have broad applicability and can augment an agent’s decision-making beyond reactive control.

Recent advances in Goal Recognition have yielded a new class of approaches capable of solving goal recognition problems without relying on predefined domain theories, defined as Model-Free Goal Recognition. Most existing approaches rely on neural networks, probabilistic theories or approximated domain theories to recognize goals without relying on explicitly defined domain knowledge. However, these approaches often neglect the causal relationships contained in the data used for their process. This oversight overlooks an opportunity to make their goal recognition process more accurate, explainable and robust. We propose a novel Model-Free Goal Recognition approach that integrates causality through Variational Inference, which, to the best of our knowledge, is an entirely novel class of techniques for goal recognition. The method encompasses three key stages: Causal Discovery, Counterfactual Inference, and decision-making grounded in Trajectory Likelihood. Our approach outperforms the existing state-of-the-art methods in all tested domains. Moreover, its strong noise resilience ensures that its performance in noisy environments is nearly indistinguishable from standard conditions, fully showcasing its robustness.

With the increasing demand for power supply reliability, efficient load restoration in large-scale distribution networks post-outage scenarios has become a critical challenge. However, traditional methods become computationally prohibitive as network expansion leads to exponential growth of decision variables. This study proposes a multi-agent reinforcement learning (MARL) framework enhanced by distribution network partitioning to address this challenge. Firstly, an improved Girvan–Newman algorithm is employed to achieve balanced partitioning of the network, defining the state space of each agent and action boundaries within the multi-agent system (MAS). Subsequently, a counterfactual reasoning framework solved by the QTRAN-alt algorithm is incorporated to refine action selection during training, thereby accelerating convergence and enhancing decision-making efficiency during execution. Experimental validation using a 27-bus system and a 70-bus system demonstrates that the proposed QTRAN-alt with the Girvan–Newman method achieves fast convergence and high returns compared to typical MARL approaches. Furthermore, the proposed methodology significantly improves the success rate of full system restoration without violating constraints.

Efficient path planning for unmanned aerial vehicles (UAVs) is crucial in remote sensing and information collection. As task scales expand, the cooperative deployment of multiple UAVs significantly improves information collection efficiency. However, collaborative communication and decision-making for multiple UAVs remain major challenges in path planning, especially in noisy environments. To efficiently accomplish complex information collection tasks in 3D space and address robust communication issues, we propose a multi-agent reinforcement learning (MARL) framework for UAV path planning based on the Counterfactual Multi-Agent Policy Gradients (COMA) algorithm. The framework incorporates attention mechanism-based UAV communication protocol and training-deployment system, significantly improving communication robustness and individual decision-making capabilities in noisy conditions. Experiments conducted on both synthetic and real-world datasets demonstrate that our method outperforms existing algorithms in terms of path planning efficiency and robustness, especially in noisy environments, achieving a 78\% improvement in entropy reduction.

This paper presents the development of an Artificial Intelligence (AI) based fighter jet agent within a customized Pygame simulation environment, designed to solve multi-objective tasks via deep reinforcement learning (DRL). The jet's primary objectives include efficiently navigating the environment, reaching a target, and selectively engaging or evading an enemy. A reward function balances these goals while optimized hyperparameters enhance learning efficiency. Results show more than 80 % task completion rate, demonstrating effective decision-making. To enhance transparency, the jet's action choices are analyzed by comparing the rewards of the actual chosen action (factual action) with those of alternate actions (counterfactual actions), providing insights into the decision-making rationale. This study illustrates DRL's potential for multi-objective problem-solving with explainable AI. Project code is available at: https://github.com/swatikar95/Autonomous-Fighter-Jet-Navigation-and-Combat.

This work suggests an adaptive, deep reinforcement learning-driven framework having five integrated modules that address important facets of volatility-aware optimization to improve the effectiveness of Test Case Prioritization (TCP). To assign priority scores using temporal and contextual attention mechanisms within a dynamic graph, the Dual-Attention Temporal Graph Prioritization Network (DAT-GPN) uses historical execution logs and changing software modifications. The Reinforcement-Driven Volatility-Aware Clustered Prioritizer (RD-VACP) uses Q-learning agents to optimize execution order and remove redundancy while clustering test cases as per volatility metrics. With the addition of epistemic and aleatoric uncertainties to a multi-agent PPO structure, the Uncertainty-Regularized Multi-Agent PPO Scheduler (UR-MAPPO) improves policy stability in dynamic test scenarios. To assess hypothetical test results for risk-aware decision-making, the Counterfactual Impact Analysis Prioritizer (CIAP) uses structural causal inference. Lastly, to balance detection time, risk exposure, and resource consumption, the Multi-Objective Adaptive Ensemble Prioritization Framework (MO-AEPF) combines reinforcement, causal, and sequential learning. This framework provides a dependable and understandable TCP solution. Dual-attention graph modeling for contextual and temporal prioritization. For risk-sensitive, optimal execution, use reinforcement and causal learning. Multi-objective ensemble optimization for resource efficiency and balanced fault detection.

Reinforcement learning has emerged as a transformative solution for optimizing bidding strategies and budget allocation in online advertising auctions. The dynamic nature of these auctions, characterized by rapid market changes and complex user behaviors, necessitates sophisticated decision-making mechanisms beyond traditional rule-based systems. By leveraging advanced machine learning techniques, including contextual bandits, Deep Q-learning networks, and actor-critic architectures, modern advertising platforms can achieve significant improvements in campaign performance and return on investment. The implementation of these systems requires careful consideration of practical challenges, including reward shaping, delayed feedback handling, and counterfactual estimation. Through effective feature engineering and model architecture optimization, these challenges can be addressed while maintaining computational efficiency and system reliability. The integration of emerging technologies, such as multi-agent systems and transfer learning, continues to push the boundaries of what's possible in automated advertising optimization, promising even greater improvements in targeting accuracy and campaign effectiveness.

The increasing complexity and interconnectedness of CPS have made them more vulnerable to dynamic failures, cyber-attacks, and performance degradation, hence requiring an autonomous self-healing capability. This paper proposes a trustworthy and adaptive Artificial Intelligence (AI) framework for autonomous self-healing in CPS by integrating Federated Learning (FL) and Explainable Artificial Intelligence (XAI) for enabling privacy-preserving, interpretable, and adaptive decision intelligence. Unlike conventional centralized approaches, the proposed framework supports the training of models in a distributed fashion across heterogeneous edge nodes while ensuring global consistency and trust via federated consensus optimization and differential privacy mechanisms. An adaptive self-healing controller dynamically reconfigures system behavior upon anomaly detection by leveraging causal reasoning and Explainable Reinforcement Learning (XRL) for transparent decision pathways. Moreover, a trust calibration layer estimates model reliability using Bayesian uncertainty quantification and graph-based trust propagation to enhance collaborative robustness among participating nodes. Exhaustive experiments on CPS benchmark datasets, covering smart grids, autonomous vehicular networks, and Industrial IoT (IIoT) environments, demonstrate fault recovery time reduction by up to 31%, falsepositive alert reduction by up to 24%, and an improvement in model interpretability metrics by up to 18% when compared to state-of-the-art baselines. The proposed FTASH framework lays a foundation for an autonomous, trustworthy, and resilient CPS, leveraging the importance of transparency and reliability in nextgeneration intelligent infrastructures.

Carrier networks High-availability carrier networks require almost zero downtime, fault isolation in less than a second, and predictive resilience of heterogeneous infrastructure across core, edge, and access layers. The models of traditional network operations, relying on the use of rule-based automation and reactive monitoring, do not suit 5G, cloud-native network functions (CNFs), and software-defined architectures. The paper suggests a system of autonomous telecommunication operations based on AI that will allow self-monitoring, self-repair, and self-optimization of carrier-grade networks. The framework proposed combines seven architectural layers: (1) multi-source ingestion of telemetry, (2) real-time stream processing, (3) anomaly detection based on the application of the deep learning or graph-based model, (4) forecasted failure based on the use of temporal neural networks, (5) root-cause analysis based on causal inference engines, (6) a closed-loop control through the use of reinforcement learning-based policy optimization, and (7) managed and conformity through explainable AI (XAI) controls. The system utilizes hybrid edge-cloud deployment in order to reduce latency and centralized intelligence coordination. Carrier-scale experimental simulations show that it reduces mean time to detect (MTTD) and mean time to repair (MTTR), increases service availability, and efficiency of resource utilization. Constant learning systems are also integrated into the framework to respond to changes in the patterns of traffic and threat. The outcomes show that autonomous operations are enabled by AI, which can change telecom environments, transforming the reactive approach to management to the predictive and prescriptive operational patterns, in favor of scalable, resilient, and cost-effective carrier networks. The model proposed is a quality direction of Level-4/Level-5 autonomous network maturity in future-generation telecom ecosystem.

End-to-end autonomous driving systems, predominantly trained through imitation learning, have demonstrated considerable effectiveness in leveraging large-scale expert driving data. Despite their success in open-loop evaluations, these systems often exhibit significant performance degradation in closed-loop scenarios due to causal confusion. This confusion is fundamentally exacerbated by the overreliance of the imitation learning paradigm on expert trajectories, which often contain unattributable noise and interfere with the modeling of causal relationships between environmental contexts and appropriate driving actions. To address this fundamental limitation, we propose Perception-Guided Self-Supervision (PGS) - a simple yet effective training paradigm that leverages perception outputs as the primary supervisory signals, explicitly modeling causal relationships in decision-making. The proposed framework aligns both the inputs and outputs of the decision-making module with perception results, such as lane centerlines and the predicted motions of surrounding agents, by introducing positive and negative self-supervision for the ego trajectory. This alignment is specifically designed to mitigate causal confusion arising from the inherent noise in expert trajectories. Equipped with perception-driven supervision, our method, built on a standard end-to-end architecture, achieves a Driving Score of 78.08 and a mean success rate of 48.64% on the challenging closed-loop Bench2Drive benchmark, significantly outperforming existing state-of-the-art methods, including those employing more complex network architectures and inference pipelines. These results underscore the effectiveness and robustness of the proposed PGS framework and point to a promising direction for addressing causal confusion and enhancing real-world generalization in autonomous driving.

End-to-end planning methods are the de-facto standard of the current autonomous driving system, while the robustness of the data-driven approaches suffers due to the notorious long-tail problem (i.e., rare but safety-critical failure cases). In this work, we explore whether recent diffusion-based video generation methods (a.k.a. world models), paired with structured 3D layouts, can enable a fully automated pipeline to self-correct such failure cases. We first introduce an agent to simulate the role of product manager, dubbed PM-Agent, which formulates data requirements to collect data similar to the failure cases. Then, we use a generative model that can simulate both data collection and annotation. However, existing generative models struggle to generate high-fidelity data conditioned on 3D layouts. To address this, we propose DriveSora, which can generate spatiotemporally consistent videos aligned with the 3D annotations requested by PM-Agent. We integrate these components into our self-correcting agentic system, CorrectAD. Importantly, our pipeline is end-to-end model agnostic and can be applied to improve any end-to-end planner. Evaluated on both nuScenes and a more challenging in-house dataset across multiple end-to-end planners, CorrectAD corrects 62.5% and 49.8% of failure cases, reducing collision rates by 39% and 27%, respectively.

Optimizing large-scale machine learning systems, such as recommendation models for global video platforms, requires navigating a massive hyperparameter search space and, more critically, designing sophisticated optimizers, architectures, and reward functions to capture nuanced user behaviors. Achieving substantial improvements in these areas is a non-trivial task, traditionally relying on extensive manual iterations to test new hypotheses. We propose a self-evolving system that leverages Large Language Models (LLMs), specifically those from Google's Gemini family, to autonomously generate, train, and deploy high-performing, complex model changes within an end-to-end automated workflow. The self-evolving system is comprised of an Offline Agent (Inner Loop) that performs high-throughput hypothesis generation using proxy metrics, and an Online Agent (Outer Loop) that validates candidates against delayed north star business metrics in live production. Our agents act as specialized Machine Learning Engineers (MLEs): they exhibit deep reasoning capabilities, discovering novel improvements in optimization algorithms and model architecture, and formulating innovative reward functions that target long-term user engagement. The effectiveness of this approach is demonstrated through several successful production launches at YouTube, confirming that autonomous, LLM-driven evolution can surpass traditional engineering workflows in both development velocity and model performance.

The safety and reliability of autonomous vehicles represent critical prerequisites for their widespread deployment, yet operational risks remain diverse and complex. This paper employs causal reasoning techniques to investigate driving risk causality in self-driving cars, aiming to identify latent risk sources and quantify their impacts on driving behavior. We propose the RF-LM (Random Forest-Linear Model) framework, a novel two-stage machine learning approach for causal effect identification and estimation, specifically addressing human-vehicle congregation scenarios with weather-related confounding factors. Validated through the AVOID autonomous vehicle incident dataset, our methodology demonstrates significant accuracy improvements in causal modeling after confounder adjustment. Comprehensive sensitivity analysis, robustness testing, and heterogeneity examination further confirm the stability and reliability of the proposed risk source model.

This article introduces a novel architecture for autonomous continuous integration and continuous deployment (CI/CD) systems capable of self-healing and self-optimization without human intervention. The article presents intelligent deployment meshes that integrate deep anomaly detection using LSTM networks with Bayesian change-point detection to identify deployment anomalies before they impact production environments. The proposed framework leverages causal CI/CD graphs to model complex interdependencies between microservices, enabling context-aware remediation strategies including automated rollbacks and intelligent canary analysis. The article's approach unifies machine learning metadata tracking (MLMD) with traditional software observability stacks, creating dual-aspect visibility that optimizes for both model-aware and application-aware pipeline configurations. The article demonstrates how semantic diffing engines can perform version-aware auto-validation, significantly reducing false positives in anomaly detection while improving remediation accuracy in multi-tenant environments. The resulting autonomous CI/CD architecture represents a paradigm shift from reactive to predictive deployment strategies, enabling organizations to maintain high availability while accelerating release velocity in complex microservice ecosystems.

In this work we aim to bridge the divide between autonomous vehicles and causal reasoning. Autonomous vehicles have come to increasingly interact with human drivers, and in many cases may pose risks to the physical or mental well-being of those they interact with. Meanwhile causal models, despite their inherent transparency and ability to offer contrastive explanations, have found limited usage within such systems. As such, we first identify the challenges that have limited the integration of structural causal models within autonomous vehicles. We then introduce a number of theoretical extensions to the structural causal model formalism in order to tackle these challenges. This augments these models to possess greater levels of modularisation and encapsulation, as well presenting temporal causal model representation with constant space complexity. We also prove through the extensions we have introduced that dynamically mutable sets (e.g. varying numbers of autonomous vehicles across time) can be used within a structural causal model while maintaining a relaxed form of causal stationarity. Finally we discuss the application of the extensions in the context of the autonomous vehicle and service robotics domain along with potential directions for future work.

Abstract: This paper presents a Real-Time Causal Self-Auditing (RCSA) framework designed to address the growing regulatory and explainability challenges associated with foundation models. By combining causal inference methods with real-time auditing pipelines, the RCSA framework delivers regulatory-grade transparency, enabling traceable and interpretable AI systems. Through synthetic but realistic experiments across multiple foundation model scales, we demonstrate significant improvements in audit latency, causal coverage, and regulatory traceability compared to baseline explainability techniques. Foundation models are increasingly deployed in sensitive and regulated domains, yet their opacity poses significant challenges for accountability and oversight. This paper introduces a novel Real-Time Causal Self-Auditing (RCSA) framework that integrates causal probing, counterfactual interventions, and adaptive monitoring layers into large foundation models. Through synthetic benchmarks and large language model case studies, we demonstrate that RCSA improves causal attribution fidelity by up to 34% compared to SHAP and Integrated Gradients, while adding only 8–12% computational overhead for models with up to 13B parameters. Keywords: Causal AI, Foundation Models, Regulatory Compliance, Model Auditing, Explainability

Recent advances in Large Language Models (LLMs) have incorporated planning and reasoning capabilities, enabling models to outline steps before execution and provide transparent reasoning paths. This enhancement has reduced errors in mathematical and logical tasks while improving accuracy. These developments have facilitated LLMs' use as agents that can interact with tools and adapt their responses based on new information. Our study examines DeepSeek R1, a model trained to output reasoning tokens similar to OpenAI's o1. Testing revealed concerning behaviors: the model exhibited deceptive tendencies and demonstrated self-preservation instincts, including attempts of self-replication, despite these traits not being explicitly programmed (or prompted). These findings raise concerns about LLMs potentially masking their true objectives behind a facade of alignment. When integrating such LLMs into robotic systems, the risks become tangible - a physically embodied AI exhibiting deceptive behaviors and self-preservation instincts could pursue its hidden objectives through real-world actions. This highlights the critical need for robust goal specification and safety frameworks before any physical implementation.

Large language models (LLMs) often have a fixed knowledge cutoff, limiting their accuracy on emerging information. We present ALAS (Autonomous Learning Agent System), a modular pipeline that continuously updates an LLM's knowledge with minimal human intervention. ALAS autonomously generates a learning curriculum for a target domain, retrieves up-to-date information from the web (with citations), distills this into question-answer training data, and fine-tunes the model through supervised fine-tuning (SFT) and direct preference optimization (DPO). It iteratively evaluates performance and revises the curriculum, enabling long-term continual learning. We demonstrate ALAS's ability to self-improve a model on rapidly evolving domains (e.g., new Python releases, latest security CVEs, academic trends), significantly boosting post-cutoff question answering accuracy (from 15% to 90% on average) without manual dataset curation. The system emphasizes modularity and reproducibility: each component (planning, retrieval, distillation, memory, fine-tuning) is interchangeable and built on standard APIs. We discuss comparative baselines (e.g., retrieval-augmented generation vs. fine-tuning) and show that ALAS achieves 90% accuracy on knowledge-updated queries with minimal engineering overhead. Finally, we outline limitations (cost, dependency on source quality) and future directions for autonomous lifelong learning in LLMs.

Effectively integrating Large Language Models (LLMs) into autonomous driving requires a balance between leveraging high-level reasoning and maintaining real-time efficiency. Existing approaches either activate LLMs too frequently, causing excessive computational overhead, or use fixed schedules, failing to adapt to dynamic driving conditions. To address these challenges, we propose AdaDrive, an adaptively collaborative slow-fast framework that optimally determines when and how LLMs contribute to decision-making. (1) When to activate the LLM: AdaDrive employs a novel adaptive activation loss that dynamically determines LLM invocation based on a comparative learning mechanism, ensuring activation only in complex or critical scenarios. (2) How to integrate LLM assistance: Instead of rigid binary activation, AdaDrive introduces an adaptive fusion strategy that modulates a continuous, scaled LLM influence based on scene complexity and prediction confidence, ensuring seamless collaboration with conventional planners. Through these strategies, AdaDrive provides a flexible, context-aware framework that maximizes decision accuracy without compromising real-time performance. Extensive experiments on language-grounded autonomous driving benchmarks demonstrate that AdaDrive state-of-the-art performance in terms of both driving accuracy and computational efficiency. Code is available at https://github.com/ReaFly/AdaDrive.

Molecular property prediction is fundamental to chemical engineering applications such as solvent screening. We present Socrates-Mol, a framework that transforms language models into empirical Bayesian reasoners through context engineering, addressing cold start problems without model fine-tuning. The system implements a reflective-prediction cycle where initial outputs serve as priors, retrieved molecular cases provide evidence, and refined predictions form posteriors, extracting reusable chemical rules from sparse data. We introduce ranking tasks aligned with industrial screening priorities and employ cross-model self-consistency across five language models to reduce variance. Experiments on amine solvent LogP prediction reveal task-dependent patterns: regression achieves 72% MAE reduction and 112% R-squared improvement through self-consistency, while ranking tasks show limited gains due to systematic multi-model biases. The framework reduces deployment costs by over 70% compared to full fine-tuning, providing a scalable solution for molecular property prediction while elucidating the task-adaptive nature of self-consistency mechanisms.

No abstract available

Large language models (LLMs) excel at complex tasks with advances in reasoning capabilities. However, existing reward mechanisms remain tightly coupled to final correctness and pay little attention to the underlying reasoning process: trajectories with sound reasoning but wrong answers receive low credit, while lucky guesses with flawed logic may be highly rewarded, affecting reasoning generalization. From a causal perspective, we interpret multi-candidate reasoning for a fixed question as a family of counterfactual experiments with theoretical supports. Building on this, we propose Group Causal Counterfactual Policy Optimization to explicitly train LLMs to learn generalizable reasoning patterns. It proposes an episodic causal counterfactual reward that jointly captures (i) robustness, encouraging the answer distribution induced by a reasoning step to remain stable under counterfactual perturbations; and (ii) effectiveness, enforcing sufficient variability so that the learned reasoning strategy can transfer across questions. We then construct token-level advantages from this reward and optimize the policy, encouraging LLMs to favor reasoning patterns that are process-valid and counterfactually robust. Extensive experiments on diverse benchmarks demonstrate its advantages.

Counterfactual reasoning, a cornerstone of human cognition and decision-making, is often seen as the 'holy grail' of causal learning, with applications ranging from interpreting machine learning models to promoting algorithmic fairness. While counterfactual reasoning has been extensively studied in contexts where the underlying causal model is well-defined, real-world causal modeling is often hindered by model and parameter uncertainty, observational noise, and chaotic behavior. The reliability of counterfactual analysis in such settings remains largely unexplored. In this work, we investigate the limitations of counterfactual reasoning within the framework of Structural Causal Models. Specifically, we empirically investigate \emph{counterfactual sequence estimation} and highlight cases where it becomes increasingly unreliable. We find that realistic assumptions, such as low degrees of model uncertainty or chaotic dynamics, can result in counterintuitive outcomes, including dramatic deviations between predicted and true counterfactual trajectories. This work urges caution when applying counterfactual reasoning in settings characterized by chaos and uncertainty. Furthermore, it raises the question of whether certain systems may pose fundamental limitations on the ability to answer counterfactual questions about their behavior.

Counterfactual reasoning, a hallmark of intelligence, consists of three steps: inferring latent variables from observations (abduction), constructing alternatives (interventions), and predicting their outcomes (prediction). This skill is essential for advancing LLMs'causal understanding and expanding their applications in high-stakes domains such as scientific research. However, existing efforts in assessing LLM's counterfactual reasoning capabilities tend to skip the abduction step, effectively reducing to interventional reasoning and leading to overestimation of LLM performance. To address this, we introduce executable counterfactuals, a novel framework that operationalizes causal reasoning through code and math problems. Our framework explicitly requires all three steps of counterfactual reasoning and enables scalable synthetic data creation with varying difficulty, creating a frontier for evaluating and improving LLM's reasoning. Our results reveal substantial drop in accuracy (25-40%) from interventional to counterfactual reasoning for SOTA models like o4-mini and Claude-4-Sonnet. To address this gap, we construct a training set comprising counterfactual code problems having if-else condition and test on out-of-domain code structures (e.g. having while-loop); we also test whether a model trained on code would generalize to counterfactual math word problems. While supervised finetuning on stronger models'reasoning traces improves in-domain performance of Qwen models, it leads to a decrease in accuracy on OOD tasks such as counterfactual math problems. In contrast, reinforcement learning induces the core cognitive behaviors and generalizes to new domains, yielding gains over the base model on both code (improvement of 1.5x-2x) and math problems. Analysis of the reasoning traces reinforces these findings and highlights the promise of RL for improving LLMs'counterfactual reasoning.

No abstract available

Portfolio managers rely on correlation-based analysis and heuristic methods that fail to capture true causal relationships driving performance. We present a hybrid framework that integrates statistical causal discovery algorithms with domain knowledge from two complementary sources: a financial knowledge graph extracted from SEC 10-K filings and large language model reasoning. Our approach systematically enhances three representative causal discovery paradigms, constraint-based (PC), score-based (GES), and continuous optimization (NOTEARS), by encoding knowledge graph constraints algorithmically and leveraging LLM conceptual reasoning for hypothesis generation. Evaluated on a synthetic financial dataset of 500 firms across 18 variables, our KG+LLM-enhanced methods demonstrate consistent improvements across all three algorithms: PC (F1: 0.622 vs. 0.459 baseline, +36%), GES (F1: 0.735 vs. 0.367, +100%), and NOTEARS (F1: 0.759 vs. 0.163, +366%). The framework enables reliable scenario analysis with mean absolute error of 0.003610 for counterfactual predictions and perfect directional accuracy for intervention effects. It also addresses critical limitations of existing methods by grounding statistical discoveries in financial domain expertise while maintaining empirical validation, providing portfolio managers with the causal foundation necessary for proactive risk management and strategic decision-making in dynamic market environments.

Large Language Models (LLMs) excel at general-purpose reasoning by leveraging broad commonsense knowledge, but they remain limited in tasks requiring personalized reasoning over multifactorial personal data. This limitation constrains their applicability in domains such as healthcare, where decisions must adapt to individual contexts. We introduce Personalized Causal Graph Reasoning, a framework that enables LLMs to reason over individual-specific causal graphs constructed from longitudinal data. Each graph encodes how user-specific factors influence targeted outcomes. In response to a query, the LLM traverses the graph to identify relevant causal pathways, rank them by estimated impact, simulate potential outcomes, and generate tailored responses. We implement this framework in the context of nutrient-oriented dietary recommendations, where variability in metabolic responses demands personalized reasoning. Using counterfactual evaluation, we assess the effectiveness of LLM-generated food suggestions for glucose control. Our method reduces postprandial glucose iAUC across three time windows compared to prior approaches. Additional LLM-as-a-judge evaluations further confirm improvements in personalization quality.

Counterfactual reasoning is widely recognized as one of the most challenging and intricate aspects of causality in artificial intelligence. In this paper, we evaluate the performance of large language models (LLMs) in counterfactual reasoning. In contrast to previous studies that primarily focus on commonsense causal reasoning, where LLMs often rely on prior knowledge for inference, we specifically assess their ability to perform counterfactual inference using a set of formal rules. To support this evaluation, we introduce a new benchmark dataset, CounterBench, comprising 1.2K counterfactual reasoning questions. The dataset is designed with varying levels of difficulty, diverse causal graph structures, distinct types of counterfactual questions, and multiple nonsensical name variants. Our experiments demonstrate that counterfactual reasoning poses a significant challenge for LLMs, with most models performing at levels comparable to random guessing. To enhance LLM's counterfactual reasoning ability, we propose a novel reasoning paradigm, CoIn, which guides LLMs through iterative reasoning and backtracking to systematically explore counterfactual solutions. Experimental results show that our method significantly improves LLM performance on counterfactual reasoning tasks and consistently enhances performance across different LLMs.

As Large Language Model (LLM) agents are increasingly tasked with high-stakes autonomous decision-making, the transparency of their reasoning processes has become a critical safety concern. While \textit{Chain-of-Thought} (CoT) prompting allows agents to generate human-readable reasoning traces, it remains unclear whether these traces are \textbf{faithful} generative drivers of the model's output or merely \textbf{post-hoc rationalizations}. We introduce \textbf{Project Ariadne}, a novel XAI framework that utilizes Structural Causal Models (SCMs) and counterfactual logic to audit the causal integrity of agentic reasoning. Unlike existing interpretability methods that rely on surface-level textual similarity, Project Ariadne performs \textbf{hard interventions} ($do$-calculus) on intermediate reasoning nodes -- systematically inverting logic, negating premises, and reversing factual claims -- to measure the \textbf{Causal Sensitivity} ($\phi$) of the terminal answer. Our empirical evaluation of state-of-the-art models reveals a persistent \textit{Faithfulness Gap}. We define and detect a widespread failure mode termed \textbf{Causal Decoupling}, where agents exhibit a violation density ($\rho$) of up to $0.77$ in factual and scientific domains. In these instances, agents arrive at identical conclusions despite contradictory internal logic, proving that their reasoning traces function as"Reasoning Theater"while decision-making is governed by latent parametric priors. Our findings suggest that current agentic architectures are inherently prone to unfaithful explanation, and we propose the Ariadne Score as a new benchmark for aligning stated logic with model action.

Large language models (LLMs) have been shown to perform better when asked to reason step-by-step before answering a question. However, it is unclear to what degree the model's final answer is faithful to the stated reasoning steps. In this paper, we perform a causal mediation analysis on twelve LLMs to examine how intermediate reasoning steps generated by the LLM influence the final outcome and find that LLMs do not reliably use their intermediate reasoning steps when generating an answer. To address this issue, we introduce FRODO, a framework to tailor small-sized LMs to generate correct reasoning steps and robustly reason over these steps. FRODO consists of an inference module that learns to generate correct reasoning steps using an implicit causal reward function and a reasoning module that learns to faithfully reason over these intermediate inferences using a counterfactual and causal preference objective. Our experiments show that FRODO significantly outperforms four competitive baselines. Furthermore, FRODO improves the robustness and generalization ability of the reasoning LM, yielding higher performance on out-of-distribution test sets. Finally, we find that FRODO's rationales are more faithful to its final answer predictions than standard supervised fine-tuning.

Counterfactual reasoning is widely recognized as one of the most challenging and intricate aspects of causality in artificial intelligence. In this paper, we evaluate the performance of large language models (LLMs) in counterfactual reasoning. In contrast to previous studies that primarily focus on commonsense causal reasoning, where LLMs often rely on prior knowledge for inference, we specifically assess their ability to perform counterfactual inference using a set of formal rules. To support this evaluation, we introduce a new benchmark dataset, CounterBench, comprising 1K counterfactual reasoning questions. The dataset is designed with varying levels of difficulty, diverse causal graph structures, distinct types of counterfactual questions, and multiple nonsensical name variants. Our experiments demonstrate that counterfactual reasoning poses a significant challenge for LLMs, with most models performing at levels comparable to random guessing. To enhance LLM's counterfactual reasoning ability, we propose a novel reasoning paradigm, CoIn, which guides LLMs through iterative reasoning and backtracking to systematically explore counterfactual solutions. Experimental results show that our method significantly improves LLM performance on counterfactual reasoning tasks and consistently enhances performance across different LLMs.Our dataset is available at https://huggingface.co/datasets/CounterBench/CounterBench.

No abstract available

Large language models show promise as autonomous decision-making agents, yet their deployment in high-stakes domains remains fraught with risk. Without architectural safeguards, LLM agents exhibit catastrophic brittleness: identical capabilities produce wildly different outcomes depending solely on prompt framing. We present Chimera, a neuro-symbolic-causal architecture that integrates three complementary components - an LLM strategist, a formally verified symbolic constraint engine, and a causal inference module for counterfactual reasoning. We benchmark Chimera against baseline architectures (LLM-only, LLM with symbolic constraints) across 52-week simulations in a realistic e-commerce environment featuring price elasticity, trust dynamics, and seasonal demand. Under organizational biases toward either volume or margin optimization, LLM-only agents fail catastrophically (total loss of \$99K in volume scenarios) or destroy brand trust (-48.6% in margin scenarios). Adding symbolic constraints prevents disasters but achieves only 43-87% of Chimera's profit. Chimera consistently delivers the highest returns (\$1.52M and \$1.96M respectively, some cases +\$2.2M) while improving brand trust (+1.8% and +10.8%, some cases +20.86%), demonstrating prompt-agnostic robustness. Our TLA+ formal verification proves zero constraint violations across all scenarios. These results establish that architectural design not prompt engineering determines the reliability of autonomous agents in production environments. We provide open-source implementations and interactive demonstrations for reproducibility.

Large language models have revolutionized agent planning by serving as the engine of heuristic guidance. However, LLM-based agents often struggle to generalize across complex environments and to adapt to stochastic feedback arising from environment–action interactions. We propose Counterfactual Planning—a method designed to improve the generalizability and adaptability of agents' actions by inferring causal representations of environmental confounders and performing counterfactual reasoning over planned actions. We formalize the agent planning process as a structural causal model, providing a mathematical formulation for causal analysis of how environmental states influence action generation and how actions affect future state transitions. To support generalizable action planning, we introduce the State Causality Evaluator (SCE), which dynamically infers task-conditioned causal representations from complex environment states; and to enhance adaptability under stochastic feedback, we propose the What-If-Not (WIN) reward, which performs counterfactual interventions to refine actions through causal evaluation. We validate our framework in an open-world environment, where experiments demonstrate improvements in both action generalization and planning adaptability.

Despite achieving high accuracy on medical benchmarks, LLMs exhibit the Einstellung Effect in clinical diagnosis--relying on statistical shortcuts rather than patient-specific evidence, causing misdiagnosis in atypical cases. Existing benchmarks fail to detect this critical failure mode. We introduce MedEinst, a counterfactual benchmark with 5,383 paired clinical cases across 49 diseases. Each pair contains a control case and a"trap"case with altered discriminative evidence that flips the diagnosis. We measure susceptibility via Bias Trap Rate--probability of misdiagnosing traps despite correctly diagnosing controls. Extensive Evaluation of 17 LLMs shows frontier models achieve high baseline accuracy but severe bias trap rates. Thus, we propose ECR-Agent, aligning LLM reasoning with Evidence-Based Medicine standard via two components: (1) Dynamic Causal Inference (DCI) performs structured reasoning through dual-pathway perception, dynamic causal graph reasoning across three levels (association, intervention, counterfactual), and evidence audit for final diagnosis; (2) Critic-Driven Graph and Memory Evolution (CGME) iteratively refines the system by storing validated reasoning paths in an exemplar base and consolidating disease-specific knowledge into evolving illness graphs. Source code is to be released.

Multimodal Large Language Models (MLLMs) have emerged as a central focus in both industry and academia, but often suffer from biases introduced by visual and language priors, which can lead to multimodal hallucination. These biases arise from the visual encoder and the Large Language Model (LLM) backbone, affecting the attention mechanism responsible for aligning multimodal inputs. Existing decoding-based mitigation methods focus on statistical correlations and overlook the causal relationships between attention mechanisms and model output, limiting their effectiveness in addressing these biases. To tackle this issue, we propose a causal inference framework termed CausalMM that applies structural causal modeling to MLLMs, treating modality priors as a confounder between attention mechanisms and output. Specifically, by employing backdoor adjustment and counterfactual reasoning at both the visual and language attention levels, our method mitigates the negative effects of modality priors and enhances the alignment of MLLM's inputs and outputs, with a maximum score improvement of 65.3% on 6 VLind-Bench indicators and 164 points on MME Benchmark compared to conventional methods. Extensive experiments validate the effectiveness of our approach while being a plug-and-play solution. Our code is available at: https://github.com/The-Martyr/CausalMM

Recent advancements in recommender systems have focused on leveraging Large Language Models (LLMs) to improve user preference modeling, yielding promising outcomes. However, current LLM-based approaches struggle to fully leverage user behavior sequences, resulting in suboptimal preference modeling for personalized recommendations. In this study, we propose a novel Counterfactual Fine-Tuning (CFT) method to address this issue by explicitly emphasizing the role of behavior sequences when generating recommendations. Specifically, we employ counterfactual reasoning to identify the causal effects of behavior sequences on model output and introduce a task that directly fits the ground-truth labels based on these effects, achieving the goal of explicit emphasis. Additionally, we develop a token-level weighting mechanism to adjust the emphasis strength for different item tokens, reflecting the diminishing influence of behavior sequences from earlier to later tokens during predicting an item. Extensive experiments on real-world datasets demonstrate that CFT effectively improves behavior sequence modeling. Our codes are available at https://github.com/itsmeyjt/CFT.

Cross-view geo-localization seeks to match geographic locations using images from varied sources, including drones and satellites. Interpreting images captured by drones poses significant challenges due to the varying positions and scales resulting from the camera’s aerial perspective. Traditional approaches have primarily focused on harnessing contextual cues, which may lead to overfitting. Therefore, it is crucial to find an optimal balance between leveraging contextual details and identifying relevant features. To address this, we introduce a novel method for cross-view geo-localization that employs counterfactual causal reasoning (CCR). This method aims to refine the model’s focus, ensuring a balanced emphasis on both the intricate details of the target structure and its broader contextual environment. Our method incorporates an Adaptive Dimension Interaction Block (ADIB), which effectively discerns feature interactions across multiple dimensions, enhanced by counterfactual causal reasoning to improve recognition of target structures and their contexts. In tasks of image-based drone-view target localization and drone navigation, our method achieves superior performance on the University-1652 and SUES-200 benchmark datasets. The code and model files will be made available at https://github.com/Cyberpunk1998/CCR.

Mental health risk is a critical global public health challenge, necessitating innovative and reliable assessment methods. With the development of large language models (LLMs), they stand out to be a promising tool for explainable mental health care applications. Nevertheless, existing approaches predominantly rely on subjective textual mental records, which can be distorted by inherent mental uncertainties, leading to inconsistent and unreliable predictions. To address these limitations, this paper introduces ProMind-LLM. We investigate an innovative approach integrating objective behavior data as complementary information alongside subjective mental records for robust mental health risk assessment. Specifically, ProMind-LLM incorporates a comprehensive pipeline that includes domain-specific pretraining to tailor the LLM for mental health contexts, a self-refine mechanism to optimize the processing of numerical behavioral data, and causal chain-of-thought reasoning to enhance the reliability and interpretability of its predictions. Evaluations of two real-world datasets, PMData and Globem, demonstrate the effectiveness of our proposed methods, achieving substantial improvements over general LLMs. We anticipate that ProMind-LLM will pave the way for more dependable, interpretable, and scalable mental health case solutions.

Tailoring persuasive conversations to users leads to more effective persuasion. However, existing dialogue systems often struggle to adapt to dynamically evolving user states. This paper presents a novel method that leverages causal discovery and counterfactual reasoning for optimizing system persuasion capability and outcomes. We employ the Greedy Relaxation of the Sparsest Permutation (GRaSP) algorithm to identify causal relationships between user and system utterance strategies, treating user strategies as states and system strategies as actions. GRaSP identifies user strategies as causal factors influencing system responses, which inform Bidirectional Conditional Generative Adversarial Networks (BiCoGAN) in generating counterfactual utterances for the system. Subsequently, we use the Dueling Double Deep Q-Network (D3QN) model to utilize counterfactual data to determine the best policy for selecting system utterances. Our experiments with the PersuasionForGood dataset show measurable improvements in persuasion outcomes using our approach over baseline methods. The observed increase in cumulative rewards and Q-values highlights the effectiveness of causal discovery in enhancing counterfactual reasoning and optimizing reinforcement learning policies for online dialogue systems.

Smart buildings generate vast streams of sensor and control data, but facility managers often lack clear explanations for anomalous energy usage. We propose InsightBuild, a two-stage framework that integrates causality analysis with a fine-tuned large language model (LLM) to provide human-readable, causal explanations of energy consumption patterns. First, a lightweight causal inference module applies Granger causality tests and structural causal discovery on building telemetry (e.g., temperature, HVAC settings, occupancy) drawn from Google Smart Buildings and Berkeley Office datasets. Next, an LLM, fine-tuned on aligned pairs of sensor-level causes and textual explanations, receives as input the detected causal relations and generates concise, actionable explanations. We evaluate InsightBuild on two real-world datasets (Google: 2017-2022; Berkeley: 2018-2020), using expert-annotated ground-truth causes for a held-out set of anomalies. Our results demonstrate that combining explicit causal discovery with LLM-based natural language generation yields clear, precise explanations that assist facility managers in diagnosing and mitigating energy inefficiencies.

Counterfactual reasoning aims at answering contrary-to-fact questions like ``Would have Alice recovered had she taken aspirin?''and corresponds to the most fine-grained layer of causation. Critically, while many counterfactual statements cannot be falsified-even by randomized experiments-they underpin fundamental concepts like individual-wise fairness. Therefore, providing models to formalize and implement counterfactual beliefs remains a fundamental scientific problem. In the Markovian setting of Pearl's causal framework, we propose an alternative approach to structural causal models to represent counterfactuals compatible with a given causal graphical model. More precisely, we introduce counterfactual models, also called canonical representations of structural causal models. They enable analysts to choose a counterfactual assumption via random-process probability distributions with preassigned marginals and characterize the counterfactual equivalence class of structural causal models. Using these representations, we present a normalization procedure to disentangle the (arbitrary and unfalsifiable) counterfactual choice from the (typically testable) interventional constraints. In contrast to structural causal models, this allows to implement many counterfactual assumptions while preserving interventional knowledge, and does not require any estimation step at the individual-counterfactual layer: only to make a choice. Finally, we illustrate the specific role of counterfactuals in causality and the benefits of our approach on theoretical and numerical examples.

Large language models (LLMs) have transformed natural language processing (NLP), enabling diverse applications by integrating large-scale pre-trained knowledge. However, their static knowledge limits dynamic reasoning over external information, especially in knowledge-intensive domains. Retrieval-Augmented Generation (RAG) addresses this challenge by combining retrieval mechanisms with generative modeling to improve contextual understanding. Traditional RAG systems suffer from disrupted contextual integrity due to text chunking and over-reliance on semantic similarity for retrieval, often resulting in shallow and less accurate responses. We propose Causal-Counterfactual RAG, a novel framework that integrates explicit causal graphs representing cause-effect relationships into the retrieval process and incorporates counterfactual reasoning grounded on the causal structure. Unlike conventional methods, our framework evaluates not only direct causal evidence but also the counterfactuality of associated causes, combining results from both to generate more robust, accurate, and interpretable answers. By leveraging causal pathways and associated hypothetical scenarios, Causal-Counterfactual RAG preserves contextual coherence, reduces hallucination, and enhances reasoning fidelity.

Causal reasoning in Large Language Models spanning association, intervention, and counterfactual inference is essential for reliable decision making in high stakes settings. As deployment shifts toward edge and resource constrained environments, quantized models such as INT8 and NF4 are becoming standard. Yet the impact of precision reduction on formal causal reasoning is poorly understood. To our knowledge, this is the first study to systematically evaluate quantization effects across all three levels of Pearls Causal Ladder. Using a 3000 sample stratified CLadder benchmark, we find that rung level accuracy in Llama 3 8B remains broadly stable under quantization, with NF4 showing less than one percent overall degradation. Interventional queries at rung 2 are the most sensitive to precision loss, whereas counterfactual reasoning at rung 3 is comparatively stable but exhibits heterogeneous weaknesses across query types such as collider bias and backdoor adjustment. Experiments on the CRASS benchmark show near identical performance across precisions, indicating that existing commonsense counterfactual datasets lack the structural sensitivity needed to reveal quantization induced reasoning drift. We further evaluate Graph Retrieval Augmented Generation using ground truth causal graphs and observe a consistent improvement in NF4 interventional accuracy of plus 1.7 percent, partially offsetting compression related degradation. These results suggest that causal reasoning is unexpectedly robust to four bit quantization, graph structured augmentation can selectively reinforce interventional reasoning, and current counterfactual benchmarks fail to capture deeper causal brittleness. This work provides an initial empirical map of compressed causal reasoning and practical guidance for deploying efficient and structurally supported causal AI systems.

Causal reasoning in event knowledge graphs is critical for explainable AI in high-stakes domains such as military intelligence and public safety. Among the various approaches to causal inference, counterfactual analysis provides a principled framework by comparing predictions on factual and counterfactual graphs, where structural interventions are applied to break spurious correlations. Recently, several GNN-based methods have been proposed for event-level causal reasoning, but most of them neglect the following facts that hinder robust causal discovery: 1) their training objectives are based on associative pattern recognition rather than intervention stability, 2) they fail to model the inherent asymmetry of causal relations, leading to invalid reverse inferences, and 3) deep GNNs suffer from over-smoothing, which dilutes causal signals in long chains. To address these issues, in this paper, we propose a novel training framework-Counterfactual Perturbation-Augmented Learning (CPAL)-that enhances causal reasoning in graph neural networks without modifying model architectures. In this framework, we introduce three semantically grounded perturbations-Edge Deletion, Entity Substitution, and Causal Inversion-to simulate structural interventions and generate counterfactual graphs. We then enforce causal invariance by minimizing a Causal Consistency Loss $(\mathcal{L}_{c f})$ that penalizes models for assigning high scores to perturbed triples. Additionally, CPAL employs a Gated RGCN to preserve high-frequency causal signals across layers. Experiments on three benchmark datasets demonstrate that CPAL outperforms state-of-the-art baselines in causal link prediction, while achieving strong robustness under distribution shifts.

LLMs suffer from critical reasoning issues such as unfaithfulness, bias, and inconsistency, since they lack robust causal underpinnings and may rely on superficial correlations rather than genuine understanding. Successive LRMs have emerged as a promising alternative, leveraging advanced training techniques such as reinforcement learning (RL) and distillation to improve task accuracy. However, the impact of these training methods on causality remains largely unexplored. In this study, we conduct a systematic causal analysis on LLMs and LRMs, examining structural causal models (SCMs) of four key variables: problem instruction (Z), thinking process (T), reasoning steps (X), and answer (Y). Our findings reveal that RLVR-trained LRMs exhibit enhanced causal reasoning capabilities, aligning more closely with ideal causal structures, while LLMs and distilled LRMs fail to address causality-related deficiencies. Our further investigation indicates that RLVR reduces spurious correlations and strengthens genuine causal patterns, thereby mitigating unfaithfulness and bias. In addition, our inspection on the dynamics of the RLVR training process observes a high correlation between reduced spurious features and improved causal structures, where the causal relationships consistently improve in the training process. This study contributes to the understanding of causality in reasoning models, highlights the critical role of RLVR in enhancing causal reasoning, and provides insights for designing future AI systems with stronger causal foundations. We release our code and data at https://github.com/Harryking1999/CoT_Causal_Analysis.

Decision making under abnormal conditions is a critical process that involves evaluating the current state and determining the optimal action to restore the system to a normal state at an acceptable cost. However, in such scenarios, existing decision-making frameworks highly rely on reinforcement learning or root cause analysis, resulting in them frequently neglecting the cost of the actions or failing to incorporate causal mechanisms adequately. By relaxing the existing causal decision framework to solve the necessary cause, we propose a minimum-cost causal decision (MiCCD) framework via counterfactual reasoning to address the above challenges. Emphasis is placed on making counterfactual reasoning processes identifiable in the presence of a large amount of mixed anomaly data, as well as finding the optimal intervention state in a continuous decision space. Specifically, it formulates a surrogate model based on causal graphs, using abnormal pattern clustering labels as supervisory signals. This enables the approximation of the structural causal model among the variables and lays a foundation for identifiable counterfactual reasoning. Once the causal structure is approximated, an optimization model is established based on counterfactual estimation. The Sequential Least Squares Programming (SLSQP) algorithm is further employed to optimize intervention strategies while taking costs into account. Experimental evaluations on both synthetic and real-world datasets reveal that MiCCD outperforms conventional methods across multiple metrics, including F1-score, cost efficiency, and ranking quality (nDCG@k values), thus validating its efficacy and broad applicability.

Large Language Models (LLMs) have performed remarkably on different tasks. The discovery of self-evolution approaches reveals that LLMs can think like humans which improves their reasoning abilities without external assistance. In this paper, we introduce a step-level framework called Self-Regeneration that can assist LLMs in deciding the consistency of each step by regenerating and giving out the confidence of responses. We select the casual reasoning domain and test Self-Regeneration on some popular LLMs. The result shows our approach improves LLMs' reasoning ability and increases the accuracy of responses.

No abstract available

Reinforcement Learning (RL) faces significant challenges in adaptive healthcare interventions, such as dementia care, where data is scarce, decisions require interpretability, and underlying patient-state dynamic are complex and causal in nature. In this work, we present a novel framework called Causal structure-aware Reinforcement Learning (CRL) that explicitly integrates causal discovery and reasoning into policy optimization. This method enables an agent to learn and exploit a directed acyclic graph (DAG) that describes the causal dependencies between human behavioral states and robot actions, facilitating more efficient, interpretable, and robust decision-making. We validate our approach in a simulated robot-assisted cognitive care scenario, where the agent interacts with a virtual patient exhibiting dynamic emotional, cognitive, and engagement states. The experimental results show that CRL agents outperform conventional model-free RL baselines by achieving higher cumulative rewards, maintaining desirable patient states more consistently, and exhibiting interpretable, clinically-aligned behavior. We further demonstrate that CRL's performance advantage remains robust across different weighting strategies and hyperparameter settings. In addition, we demonstrate a lightweight LLM-based deployment: a fixed policy is embedded into a system prompt that maps inferred states to actions, producing consistent, supportive dialogue without LLM finetuning. Our work illustrates the promise of causal reinforcement learning for human-robot interaction applications, where interpretability, adaptiveness, and data efficiency are paramount.

: It has become clear that mere correlations extracted from data through statistical processes are insufficient to give insight into the causal relationships inherent in them. Causal models support the necessary understanding of these relationships to make transparent and robust decisions. In a distributed setting, the causal models that are shared between agents improve their coordination and collaboration. They learn individually and from each other to optimise a system’s behaviour. We propose a combination of causal models and multi-agent reinforcement learning to create reliable and trustworthy AI systems. This combination strengthens the modelling and reasoning of agents that communicate and collaborate using shared causal insights. A comprehensive method for applying and integrating these aspects is being developed.

Renewable energy is gradually integrating into modern power systems, but its inherent volatility and uncertainty pose significant challenges to real-time scheduling and operational reliability. Traditional reinforcement learning (RL) methods are limited by their limited generalization ability and often cannot adapt to scenarios not encountered during training. To address these limitations, this paper proposes a new framework that combines causal reinforcement learning with an attention mechanism. In this method, the scheduling problem is formulated as a Markov decision process, and both basic rewards and causal rewards are incorporated to guide the learning process. The attention mechanism dynamically updates the weights of each operational target, enabling the agent to focus on key elements and effectively respond to changing environmental conditions. Comprehensive experiments were conducted using actual data from the UK power system in 2020, covering all seasonal variations and a wide range of operational scenarios. The proposed method was systematically evaluated under various conditions, including extreme and unseen cases. The results show that this framework significantly enhances the adaptability, robustness, and generalization performance of reinforcement learning algorithms. Compared to traditional methods, it achieves higher operational efficiency and renewable energy utilization. These findings highlight the potential of causal relationships and attention-based reinforcement learning in achieving intelligent, reliable, and scalable energy management in future power systems.

The residual stress field in die-forged blanks exhibits complex distribution patterns and high magnitudes, serving as the primary cause of significant milling distortion in aluminum alloy structural components. To mitigate this issue, finishing allowances are introduced to control milling distortion. Variations in residual stress fields across blanks necessitate online decision-making for distortion control. Although data-driven methods theoretically enable such decision-making, traditional approaches are hampered by imbalanced training data and confounding factors, making it difficult to establish stable relationships among variables, and limiting the robustness of the model’s decisions. To overcome this challenge, this study explores the underlying mechanism of distortion control via finishing allowance. The proposed Causal Deep Q-Network (CDQN) method constructs an intelligent agent based on a causal graph. The residual stress field of the blank is used as an instrumental variable, and intervention operations are applied to block the backdoor path, enabling the identification of causal effects among variables. Verification tests were conducted on typical structural components (600 mm × 240 mm × 27 mm) using 7075-T6 die-forged blanks. The average distortion observed was 0.027 mm (variance: 9.5 × 10 −6 ) in simulation, and 0.077 mm (variance: 7 × 10 −3 ) in actual milling tests on three components. These results validate the proposed method’s effectiveness in precisely controlling milling distortion in structural components.

The iterated prisoner’s dilemma (IPD) is an archetypal paradigm to model cooperation and has guided studies on social dilemmas. In this work, we develop a causal reinforcement learning (CRL) strategy in a PD game. An agent is designed to have an explicit causal representation of other agents playing strategies from the Axelrod tournament. The collection of policies is assembled in an ensemble RL to choose the best strategy. The agent is then tested against selected Axelrod tournament strategies as well as an adaptive agent trained using traditional RL. Results show that our agent is able to play against all other players and score higher while being adaptive in situations where the strategy of the other players’ changes. Furthermore, the decision taken by the agent can be explained in terms of the causal representation of the interactions. Based on the decision made by the agent, a human observer can understand the chosen strategy.

In causal reinforcement learning (RL), counterfactual reasoning deals with “what if” situations and allows for investigating the potential consequences of actions or events that did not actually happen. In this paper, we combine counterfactual reasoning and reinforcement learning (RL) and propose Counterfactually-Guided Causal Reinforcement Learning with Reward Machines (CGC-RL). In CGC-RL, using observational data, we first compute the optimal counterfactual sequence with the highest probability of completing a given task. Then, we construct an RM compatible with the counterfactual sequence. We use the constructed RM to apply dynamic potential-based reward shaping to encourage the agent to follow the counterfactual sequence. We prove the policy-invariance under dynamic reward shaping with RMs. Finally, we implement CGC-RL in one case study and compare the results with three baselines. Our results show that CGC-RL outperforms the baselines.

[Context] Multi-agent reinforcement learning (MARL) has shown strong performance in environments requiring coordinated behavior. However, transferring knowledge remains a significant challenge in dynamic settings with changing goals. [Problem] Traditional knowledge transfer methods in MARL struggle to generalize, and agents often require costly retraining to adapt. [Approach] This paper introduces a causal knowledge transfer framework that enables agents to learn and share compact causal representations of paths within a non-stationary environment. As the environment changes (with new obstacles), agents' collisions necessitate adaptive recovery strategies. We model each collision as a causal intervention instantiated as a sequence of recovery action macros whose effect corresponds to causal knowledge of how to circumvent the obstacle while increasing the chances of achieving the agent's goal (maximizing cumulative reward). The recovery action macro is transferred online from a teacher agent in a zero-shot fashion, i.e., without retraining, just by querying a lookup model with local context information (collisions). [Results] Our findings reveal two key insights: (1) agents with heterogeneous goals were able to bridge about half of the gap between random exploration and a fully retrained policy when adapting to new environments, and (2) the effectiveness of knowledge transfer depends on the interplay between environment complexity and heterogeneity of agents' goals.

Since the advent of autonomous driving technology, it has experienced remarkable progress over the last decade. However, most existing research still struggles to address the challenges posed by environments where multiple vehicles have to interact seamlessly. This study aims to integrate causal learning with reinforcement learning-based methods by leveraging causal disentanglement representation learning (CDRL) to identify and extract causal features that influence optimal decision-making in autonomous vehicles. These features are then incorporated into graph neural network-based reinforcement learning algorithms to enhance decision-making in complex traffic scenarios. By using causal features as inputs, the proposed approach enables the optimization of vehicle behavior at an unsignalized intersection. Experimental results demonstrate that our proposed method achieves the highest average reward during training and our approach significantly outperforms other learning-based methods in several key metrics such as collision rate and average cumulative reward during testing. This study provides a promising direction for advancing multi-agent autonomous driving systems and make autonomous vehicles'navigation safer and more efficient in complex traffic environments.

Dynamic service migration is essential for guaranteeing quality of service (QoS) for mobile users in multi-access edge computing (MEC). However, conventional reinforcement learning and other strategies frequently result in suboptimal policies. These methods are unable to differentiate between actual cause-and-effect relationships and fake patterns due to their reliance on correlational data, resulting in inefficient and expensive migrations. This paper introduces a novel digital twinassisted causal multi-agent reinforcement learning (DT-CausalMARL) framework to address this fundamental limitation. We change the optimization objective from minimizing direct system cost to maximizing the causal gain of migration actions. This is accomplished by employing a digital twin as a counterfactual analysis engine to assess the actual consequences of decisions in comparison to a baseline policy. This causal gain is subsequently employed as a robust reward signal by a causal multi-agent deep deterministic policy gradient (Causal-MADDPG) algorithm to train cooperative agents. Our framework outperforms standard MARL baselines by reducing harmful ‘regretful migrations' and improving system stability under dynamic traffic loads.

No abstract available

No abstract available

Dynamic path planning enables vehicles to autonomously navigate in unknown or continuously changing environments, thereby reducing reliance on fixed maps. Deep reinforcement learning (DRL), with its superior performance in handling high-dimensional state spaces and complex dynamic environments, has been widely applied to dynamic path planning. Traditional DRL methods are prone to capturing unnecessary noise information and irrelevant features during the training process, leading to instability and decreased adaptability of models in complex dynamic environments. To address this challenge, we propose a dynamic path-planning method based on our Causal State-Masking Twin-delayed Deep Deterministic Policy Gradient (CSM-TD3) algorithm. CSM-TD3 integrates a causal inference mechanism by introducing dynamic state masks and intervention mechanisms, allowing the policy network to focus on genuine causal features for decision optimization and thereby enhancing the convergence speed and generalization capabilities of the agent. Furthermore, causal state-masking DRL allows the system to learn the optimal mask configurations through backpropagation, enabling the model to adaptively adjust the causal features of interest. Extensive experimental results demonstrate that this method significantly enhances the convergence of the TD3 algorithm and effectively improves its performance in dynamic path planning.

No abstract available

No abstract available

We introduce State Abstraction with Causal dy-namics and Reward Machine learning (SACReM), a framework that accelerates reinforcement learning (RL) by incorporating causal knowledge through state abstraction. As specialized Mealy machines, reward machines effectively represent complex non-Markovian reward functions in RL environments. Nevertheless, these machines typically include redundant states and transitions that create unnecessary dependencies among state variables and actions that can be eliminated. By identifying and removing these superfluous elements, our approach enables RL agents to discover causal relationships through environmental exploration and refine policies via reward machines. Experimental results show that SAC ReM significantly improves agent capabilities in environments characterized by sparse and delayed rewards. We validate our method through two detailed case studies, which confirm that our algorithm successfully identifies causal structures, effectively implements reward machines in complex scenarios, and outperforms existing causal RL approaches.

Why do reinforcement learning (RL) policies fail or succeed? This is a challenging question due to the complex, high-dimensional nature of agent-environment interactions. In this work, we take a causal perspective on explaining the behavior of RL policies by viewing the states, actions, and rewards as variables in a low-level causal model. We introduce random perturbations to policy actions during execution and observe their effects on the cumulative reward, learning a simplified high-level causal model that explains these relationships. To this end, we develop a nonlinear Causal Model Reduction framework that ensures approximate interventional consistency, meaning the simplified high-level model responds to interventions in a similar way as the original complex system. We prove that for a class of nonlinear causal models, there exists a unique solution that achieves exact interventional consistency, ensuring learned explanations reflect meaningful causal patterns. Experiments on both synthetic causal models and practical RL tasks-including pendulum control and robot table tennis-demonstrate that our approach can uncover important behavioral patterns, biases, and failure modes in trained RL policies.

The dependency on extensive expert knowledge for defining subgoals in hierarchical reinforcement learning (HRL) restricts the training efficiency and adaptability of HRL agents in complex, dynamic environments. Inspired by human-guided causal discovery skills, we proposed a novel method, Human Causal Perception and Inference-driven Hierarchical Reinforcement Learning (HCPI-HRL), designed to infer diverse, effective subgoal structures as intrinsic rewards and incorporate critical objects from dynamic environmental states using stable causal relationships. The HCPI-HRL method is supposed to guide an agent's exploration direction and promote the reuse of learned subgoal structures across different tasks. Our designed HCPI-HRL comprises two levels: the top level operates as a meta controller, assigning subgoals discovered based on human-driven causal critical object perception and causal structure inference; the bottom level employs the Proximal Policy Optimisation (PPO) algorithm to accomplish the assigned subgoals. Experiments conducted across discrete and continuous control environments demonstrated that HCPI-HRL outperforms benchmark methods such as hierarchical and adjacency PPO in terms of training efficiency, exploration capability, and transferability. Our research extends the potential of HRL methods incorporating human-guided causal modelling to infer the effective relationships across subgoals, enhancing the agent's capability to learn efficient policies in dynamic environments with sparse reward signals.

Explaining agent's decision can offer valuable insights for designers and end-users. One proposed method for explaining an agent's decision-making involves representing a causal relation between state components and action as a causal model and providing explanations for the decisions made using causal model. However, traditional causal model often faces structural limitation, restricting the range of representable control problems. Additionally, providing accurate explanation becomes challenging in control problems with various types of rewards because agent's intention of an action is unknown. In this study, we introduce a causal model capable of representing a broader range of control problems and a method to provide accurate explanations in control problems with various types of reward structures. Through redefining the relationships between nodes in the causal model, we have enabled a broader representation of control problems. Also, by incorporating agent's intention into the explanation, we have achieved to provide a more precise explanation. To validate the effectiveness of our proposed method, we conducted experiments using OpenAI's LunarLander environment. Using a proposed causal model, we defined the causal model of LunarLander, which could not be represented by conventional causal models. Furthermore, by incorporating the intentions of an agent into the explanation, novel interpretations previously inaccessible have become feasible.

Causal inference is crucial for humans to explore the world, which can be modeled to enable an agent to efficiently explore the environment in reinforcement learning. Existing research indicates that establishing the causality between action and state transition will enhance an agent to reason how a policy affects its future trajectory, thereby promoting directed exploration. However, it is challenging to measure the causality due to its intractability in the vast state-action space of complex scenarios. In this paper, we propose a novel Goal Discovery with Causal Capacity (GDCC) framework for efficient environment exploration. Specifically, we first derive a measurement of causality in state space, \emph{i.e.,} causal capacity, which represents the highest influence of an agent's behavior on future trajectories. After that, we present a Monte Carlo based method to identify critical points in discrete state space and further optimize this method for continuous high-dimensional environments. Those critical points are used to uncover where the agent makes important decisions in the environment, which are then regarded as our subgoals to guide the agent to make exploration more purposefully and efficiently. Empirical results from multi-objective tasks demonstrate that states with high causal capacity align with our expected subgoals, and our GDCC achieves significant success rate improvements compared to baselines.

Intelligent decision-making within large and redundant action spaces remains challenging in deep reinforcement learning. Considering similar but ineffective actions at each step can lead to repetitive and unproductive trials. Existing methods attempt to improve agent exploration by reducing or penalizing redundant actions, yet they fail to provide quantitative and reliable evidence to determine redundancy. In this paper, we propose a method to improve exploration efficiency by estimating the causal effects of actions. Unlike prior methods, our approach offers quantitative results regarding the causality of actions for one-step transitions. We first pre-train an inverse dynamics model to serve as prior knowledge of the environment. Subsequently, we classify actions across the entire action space at each time step and estimate the causal effect of each action to suppress redundant actions during exploration. We provide a theoretical analysis to demonstrate the effectiveness of our method and present empirical results from simulations in environments with redundant actions to evaluate its performance. Our implementation is available at https://github.com/agi-brain/cee.git.

Designing effective task sequences is crucial for curriculum reinforcement learning (CRL), where agents must gradually acquire skills by training on intermediate tasks. A key challenge in CRL is to identify tasks that promote exploration, yet are similar enough to support effective transfer. While recent approach suggests comparing tasks via their Structural Causal Models (SCMs), the method requires access to ground-truth causal structures, an unrealistic assumption in most RL settings. In this work, we propose Causal-Paced Deep Reinforcement Learning (CP-DRL), a curriculum learning framework aware of SCM differences between tasks based on interaction data approximation. This signal captures task novelty, which we combine with the agent's learnability, measured by reward gain, to form a unified objective. Empirically, CP-DRL outperforms existing curriculum methods on the Point Mass benchmark, achieving faster convergence and higher returns. CP-DRL demonstrates reduced variance with comparable final returns in the Bipedal Walker-Trivial setting, and achieves the highest average performance in the Infeasible variant. These results indicate that leveraging causal relationships between tasks can improve the structure-awareness and sample efficiency of curriculum reinforcement learning. We provide the full implementation of CP-DRL to facilitate the reproduction of our main results at https://github.com/Cho-Geonwoo/CP-DRL.

Despite the advantages of multi-agent reinforcement learning (MARL) for wireless use case such as medium access control (MAC), their real-world deployment in Internet of Things (IoT) is hindered by their sample inefficiency. To alleviate this challenge, one can leverage model-based reinforcement learning (MBRL) solutions, however, conventional MBRL approaches rely on black-box models that are not interpretable and cannot reason. In contrast, in this paper, a novel causal model-based MARL framework is developed by leveraging tools from causal learn- ing. In particular, the proposed model can explicitly represent causal dependencies between network variables using structural causal models (SCMs) and attention-based inference networks. Interpretable causal models are then developed to capture how MAC control messages influence observations, how transmission actions determine outcomes, and how channel observations affect rewards. Data augmentation techniques are then used to generate synthetic rollouts using the learned causal model for policy optimization via proximal policy optimization (PPO). Analytical results demonstrate exponential sample complexity gains of causal MBRL over black-box approaches. Extensive simulations demonstrate that, on average, the proposed approach can reduce environment interactions by 58%, and yield faster convergence compared to model-free baselines. The proposed approach inherently is also shown to provide interpretable scheduling decisions via attention-based causal attribution, revealing which network conditions drive the policy. The resulting combination of sample efficiency and interpretability establishes causal MBRL as a practical approach for resource-constrained wireless systems.

Steering cooperative multi-agent reinforcement learning (MARL) towards desired outcomes is challenging, particularly when the global guidance from a human on the whole multi-agent system is impractical in a large-scale MARL. On the other hand, designing external mechanisms (e.g., intrinsic rewards and human feedback) to coordinate agents mostly relies on empirical studies, lacking a easy-to-use research tool. In this work, we employ multi-agent influence diagrams (MAIDs) as a graphical framework to address the above issues. First, we introduce the concept of MARL interaction paradigms (orthogonal to MARL learning paradigms), using MAIDs to analyze and visualize both unguided self-organization and global guidance mechanisms in MARL. Then, we design a new MARL interaction paradigm, referred to as the targeted intervention paradigm that is applied to only a single targeted agent, so the problem of global guidance can be mitigated. In implementation, we introduce a causal inference technique, referred to as Pre-Strategy Intervention (PSI), to realize the targeted intervention paradigm. Since MAIDs can be regarded as a special class of causal diagrams, a composite desired outcome that integrates the primary task goal and an additional desired outcome can be achieved by maximizing the corresponding causal effect through the PSI. Moreover, the bundled relevance graph analysis of MAIDs provides a tool to identify whether an MARL learning paradigm is workable under the design of an MARL interaction paradigm. In experiments, we demonstrate the effectiveness of our proposed targeted intervention, and verify the result of relevance graph analysis.

To reliably deploy Multi-Agent Reinforcement Learning (MARL) systems, it is crucial to understand individual agent behaviors. While prior work typically evaluates overall team performance based on explicit reward signals, it is unclear how to infer agent contributions in the absence of any value feedback. In this work, we investigate whether meaningful insights into agent behaviors can be extracted solely by analyzing the policy distribution. Inspired by the phenomenon that intelligent agents tend to pursue convergent instrumental values, we introduce Intended Cooperation Values (ICVs), a method based on information-theoretic Shapley values for quantifying each agent's causal influence on their co-players'instrumental empowerment. Specifically, ICVs measure an agent's action effect on its teammates'policies by assessing their decision (un)certainty and preference alignment. By analyzing action effects on policies and value functions across cooperative and competitive MARL tasks, our method identifies which agent behaviors are beneficial to team success, either by fostering deterministic decisions or by preserving flexibility for future action choices, while also revealing the extent to which agents adopt similar or diverse strategies. Our proposed method offers novel insights into cooperation dynamics and enhances explainability in MARL systems.

Causal discovery is the challenging task of inferring causal structure from data. Motivated by Pearl's Causal Hierarchy (PCH), which tells us that passive observations alone are not enough to distinguish correlation from causation, there has been a recent push to incorporate interventions into machine learning research. Reinforcement learning provides a convenient framework for such an active approach to learning. This paper presents CORE, a deep reinforcement learning-based approach for causal discovery and intervention planning. CORE learns to sequentially reconstruct causal graphs from data while learning to perform informative interventions. Our results demonstrate that CORE generalizes to unseen graphs and efficiently uncovers causal structures. Furthermore, CORE scales to larger graphs with up to 10 variables and outperforms existing approaches in structure estimation accuracy and sample efficiency. All relevant code and supplementary material can be found at https://github.com/sa-and/CORE.

Semi-supervised anomaly detection for sensor signals is critical in ensuring system reliability in smart manufacturing. However, existing methods rely heavily on data correlation, neglecting causality and leading to potential misinterpretations due to confounding factors. Moreover, while current reinforcement learning-based methods can effectively identify known and unknown anomalies with limited labeled samples, these methods still face several challenges, such as under-utilization of priori knowledge, lack of model flexibility, and deficient reward feedback during environmental interactions. To address the above problems, this paper innovatively constructs a counterfactual causal reinforcement learning model, termed Triple-Assisted Causal Reinforcement Learning Anomaly Detector (Tri-CRLAD). The model leverages causal inference to extract the intrinsic causal feature in data, enhancing the agent's utilization of prior knowledge and improving its generalization capability. In addition, Tri-CRLAD features a triple decision support mechanism, including a sampling strategy based on historical similarity, an adaptive threshold smoothing adjustment strategy, and an adaptive decision reward mechanism. These mechanisms further enhance the flexibility and generalization ability of the model, enabling it to effectively respond to various complex and dynamically changing environments. Experimental results across seven diverse sensor signal datasets demonstrate that Tri-CRLAD outperforms nine state-of-the-art baseline methods. Notably, Tri-CRLAD achieves up to a 23\% improvement in anomaly detection stability with minimal known anomaly samples, highlighting its potential in semi-supervised anomaly detection scenarios. Our code is available at https://github.com/Aoudsung/Tri-CRLAD.

Modern reinforcement learning (RL) struggles to capture real-world cause-and-effect dynamics, leading to inefficient exploration due to extensive trial-and-error actions. While recent efforts to improve agent exploration have leveraged causal discovery, they often make unrealistic assumptions of causal variables in the environments. In this paper, we introduce a novel framework, Variable-Agnostic Causal Exploration for Reinforcement Learning (VACERL), incorporating causal relationships to drive exploration in RL without specifying environmental causal variables. Our approach automatically identifies crucial observation-action steps associated with key variables using attention mechanisms. Subsequently, it constructs the causal graph connecting these steps, which guides the agent towards observation-action pairs with greater causal influence on task completion. This can be leveraged to generate intrinsic rewards or establish a hierarchy of subgoals to enhance exploration efficiency. Experimental results showcase a significant improvement in agent performance in grid-world, 2d games and robotic domains, particularly in scenarios with sparse rewards and noisy actions, such as the notorious Noisy-TV environments.

Reinforcement learning (RL) and causal modelling naturally complement each other. The goal of causal modelling is to predict the effects of interventions in an environment, while the goal of reinforcement learning is to select interventions that maximize the rewards the agent receives from the environment. Reinforcement learning includes the two most powerful sources of information for estimating causal relationships: temporal ordering and the ability to act on an environment. This paper examines which reinforcement learning settings we can expect to benefit from causal modelling, and how. In online learning, the agent has the ability to interact directly with their environment, and learn from exploring it. Our main argument is that in online learning, conditional probabilities are causal, and therefore offline RL is the setting where causal learning has the most potential to make a difference. Essentially, the reason is that when an agent learns from their {\em own} experience, there are no unobserved confounders that influence both the agent's own exploratory actions and the rewards they receive. Our paper formalizes this argument. For offline RL, where an agent may and typically does learn from the experience of {\em others}, we describe previous and new methods for leveraging a causal model, including support for counterfactual queries.

In this work, we propose a novel algorithmic framework for data sharing and coordinated exploration for the purpose of learning more data-efficient and better performing policies under a concurrent reinforcement learning (CRL) setting. In contrast to other work which make the assumption that all agents act under identical environments, we relax this restriction and instead consider the formulation where each agent acts within an environment which shares a global structure but also exhibits individual variations. Our algorithm leverages a causal inference algorithm in the form of Additive Noise Model - Mixture Model (ANM-MM) in extracting model parameters governing individual differentials via independence enforcement. We propose a new data sharing scheme based on a similarity measure of the extracted model parameters and demonstrate superior learning speeds on a set of autoregressive, pendulum and cart-pole swing-up tasks and finally, we show the effectiveness of diverse action selection between common agents under a sparse reward setting. To the best of our knowledge, this is the first work in considering non-identical environments in CRL and one of the few works which seek to integrate causal inference with reinforcement learning (RL).

Exploration is crucial for deep reinforcement learning and has garnered significant attention. In multi-agent environments, exploration becomes especially challenging due to the increased complexity of agent interactions and state space. However, existing exploration methods for multi-agent reinforcement learning (MARL) still cannot effectively identify states worth exploring, resulting in a large amount of inefficient exploration. In this work, we propose Causality-Guided multiagent Exploration (CGE), a novel framework that enhances multi-agent exploration by leveraging the causal relation between the agents and the environment. The key insight of CGE is that exploration becomes more effective when agents understand how their actions influence the environment. To this end, we introduce a state-dependent measure of causal influence based on conditional average treatment effect and demonstrate that it reliably guides agents to take impactful actions. Experimental results demonstrate that our approach achieves superior performance compared to state-of-the-art MARL baselines across various tasks in both Google Research Football (GRF) and the StarCraft Multi-Agent Challenge (SMAC).

Understanding the agent's learning process, particularly the factors that contribute to its success or failure post-training, is crucial for comprehending the rationale behind the agent's decision-making process. Prior methods clarify the learning process by creating a structural causal model (SCM) or visually representing the distribution of value functions. Nevertheless, these approaches have constraints as they exclusively function in 2D-environments or with uncomplicated transition dynamics. Understanding the agent's learning process in complicated environments or tasks is more challenging. In this paper, we propose REVEAL-IT, a novel framework for explaining the learning process of an agent in complex environments. Initially, we visualize the policy structure and the agent's learning process for various training tasks. By visualizing these findings, we can understand how much a particular training task or stage affects the agent's performance in test. Then, a GNN-based explainer learns to highlight the most important section of the policy, providing a more clear and robust explanation of the agent's learning process. The experiments demonstrate that explanations derived from this framework can effectively help in the optimization of the training tasks, resulting in improved learning efficiency and final performance.

The proliferation of agentic coding systems-software that writes software-has precipitated a structural lacuna in financial supervision. This Ordo-Causal Attribution Deficit describes the inability of existing prudential frameworks to assign legal duty, supervisory jurisdiction, and liability in chains of automated code generation that extend across days, multiple agentic nodes, and organizational boundaries. As the 2026 Agentic Coding Trends Report demonstrates, financial institutions now deploy "long-running agents" capable of building complete systems over extended periods with "minimal human intervention," while multi-agent orchestration creates emergent behaviours that defy linear causation. Current regulatory architectures assume proximate human oversight and discrete, attributable acts of programming. When autonomous agents generate code that modifies core banking systems, creates financial algorithms, or deploys customer-facing interfaces, the traditional nexus between human intent, supervisory duty, and prudential outcome fractures. This framework addresses the deficit by establishing forensic standards for agentic causation, a quantifiable risk assessment protocol, and capital-based incentives that internalize the externalities of unsupervised machine generation in financial services.

Pre-trained Vision-Language-Action (VLA) models represent a major leap towards general-purpose robots, yet efficiently adapting them to novel, specific tasks in-situ remains a significant hurdle. While reinforcement learning (RL) is a promising avenue for such adaptation, the process often suffers from low efficiency, hindering rapid task mastery. We introduce Reflective Self-Adaptation, a framework for rapid, autonomous task adaptation without human intervention. Our framework establishes a self-improving loop where the agent learns from its own experience to enhance both strategy and execution. The core of our framework is a dual-pathway architecture that addresses the full adaptation lifecycle. First, a Failure-Driven Reflective RL pathway enables rapid learning by using the VLM's causal reasoning to automatically synthesize a targeted, dense reward function from failure analysis. This provides a focused learning signal that significantly accelerates policy exploration. However, optimizing such proxy rewards introduces a potential risk of"reward hacking,"where the agent masters the reward function but fails the actual task. To counteract this, our second pathway, Success-Driven Quality-Guided SFT, grounds the policy in holistic success. It identifies and selectively imitates high-quality successful trajectories, ensuring the agent remains aligned with the ultimate task goal. This pathway is strengthened by a conditional curriculum mechanism to aid initial exploration. We conduct experiments in challenging manipulation tasks. The results demonstrate that our framework achieves faster convergence and higher final success rates compared to representative baselines. Our work presents a robust solution for creating self-improving agents that can efficiently and reliably adapt to new environments.

Despite significant progress, recent studies indicate that current large language models (LLMs) may still capture dataset biases and utilize them during inference, leading to the poor generalizability of LLMs. However, due to the diversity of dataset biases and the insufficient nature of bias suppression based on in-context learning, the effectiveness of previous prior knowledge-based debiasing methods and in-context learning based automatic debiasing methods is limited. To address these challenges, we explore the combination of causal mechanisms with information theory and propose an information gain-guided causal intervention debiasing (ICD) framework. To eliminate biases within the instruction-tuning dataset, it is essential to ensure that these biases do not provide any additional information to predict the answers, i.e., the information gain of these biases for predicting the answers needs to be 0. Under this guidance, this framework utilizes a causal intervention-based data rewriting method to automatically and autonomously balance the distribution of instruction-tuning dataset for reducing the information gain. Subsequently, it employs a standard supervised fine-tuning process to train LLMs on the debiased dataset. Experimental results show that ICD can effectively debias LLM to improve its generalizability across different tasks.

No abstract available

We introduce DriveAgent, a modular multi-agent autonomous driving framework that leverages large language model (LLM) reasoning combined with multimodal sensor fusion for autonomous driving. DriveAgent orchestrates specialized agents operating on camera, Light Detection and Ranging (LiDAR), Inertial Measurement Unit (IMU), and Global Positioning System (GPS) with LLM-driven analytical processes to deliver temporally aligned perception, causal reasoning, and action recommendations. The framework operates through a modular agent-based pipeline comprising four principal modules: (i) a descriptive analysis agent identifying critical sensor data events based on filtered timestamps, (ii) dedicated vehicle-level analysis conducted by LiDAR and vision agents that collaboratively assess vehicle conditions and movements, (iii) environmental reasoning and causal analysis agents explaining contextual changes and their underlying mechanisms, and (iv) an urgency-aware decision-generation agent prioritizing insights and proposing timely maneuvers. This modular design empowers the LLM to effectively coordinate specialized perception and reasoning agents, delivering cohesive, interpretable insights into complex autonomous driving scenarios. Extensive experiments demonstrate that DriveAgent substantially outperforms baseline methods, achieving a 26.31% improvement in vehicle reasoning and consistent enhancements of up to 2.85% in environmental reasoning. These results highlight the effectiveness of our LLM-driven multi-agent sensor fusion framework in boosting the robustness and reliability of autonomous driving systems.

Target-driven visual navigation presents great potentials in scientific and industrial fields. It takes the target and environment observations as input. However, during training, we found that the agent sometimes got stuck in specific locations. Based on the analysis on visual information from a novel causal perspective, one of the most critical hurdles is the neglect of confounders in environments, which often leads to spurious correlations. Mitigating the confounding effect helps to discover the real causality and therefore are taken into consideration in other fields such as object detection. In this article, we propose Causal Intervention Visual Navigation (CIVN), based on deep reinforcement learning (DRL) and causal intervention. We realize causal intervention using front-door adjustment as most confounders are hard to model explicitly. Specifically, CIVN is implemented by Causal Attention, which is a reasonable approximation of causal intervention for visual navigation. Causal attention provides high-quality representation, which is leveraged by DRL and reduces the number of “stuck”. It is worth mentioned that causal intervention is for the first time applied by us in solving the confounding effect in target-driven visual navigation. Extensive experiments on AI2-THOR demonstrate that CIVN achieves better performance than prior arts. Specifically, the generalization for unknown targets and scenes is improved by a large margin, which is a basic research topic in visual navigation. Moreover, to obtain better generalization, we propose a novel experiment utilizing pre-trained models firstly.

Autonomous driving systems often encounter long-tail corner cases that are rare, ambiguous, and not well represented in training data. Traditional deep learning models struggle to generalize in these cases, as they rely heavily on pattern recognition and lack commonsense reasoning. We propose an end-to-end framework that enhances visual perception through causal intervention and guides behavior decision-making using Multimodal Large Language Models (MLLMs). First, a causality-aware object detection module is introduced to reduce contextual bias by learning interventional representations. Then, visual and historical state information is encoded into structured prompts, enabling MLLMs to perform natural language-based reasoning and generate interpretable driving commands. Experimental results on multiple complex scenarios show that our method improves generalization, robustness, and decision consistency compared to conventional baselines.

Transparency and explainability are important features that responsible autonomous vehicles should possess, particularly when interacting with humans, and causal reasoning offers a strong basis to provide these qualities. However, even if one assumes agents act to maximise some concept of reward, it is difficult to make accurate causal inferences of agent planning without capturing what is of importance to the agent. Thus our work aims to learn a weighting of reward metrics for agents such that explanations for agent interactions can be causally inferred. We validate our approach quantitatively and qualitatively across three real-world driving datasets, demonstrating a functional improvement over previous methods and competitive performance across evaluation metrics.

As autonomous agents powered by large language models (LLMs) continue to demonstrate potential across various assistive tasks, ensuring their safe and reliable behavior is crucial for preventing unintended consequences. In this work, we introduce CIP, a novel technique that leverages causal influence diagrams (CIDs) to identify and mitigate risks arising from agent decision-making. CIDs provide a structured representation of cause-and-effect relationships, enabling agents to anticipate harmful outcomes and make safer decisions. Our approach consists of three key steps: (1) initializing a CID based on task specifications to outline the decision-making process, (2) guiding agent interactions with the environment using the CID, and (3) iteratively refining the CID based on observed behaviors and outcomes. Experimental results demonstrate that our method effectively enhances safety in both code execution and mobile device control tasks.

No abstract available

Behavior analysis across species represents a fundamental challenge in neuroscience, psychology, and ethology, typically requiring extensive expert knowledge and labor-intensive processes that limit research scalability and accessibility. We introduce BehaveAgent, an autonomous multimodal AI agent designed to automate behavior analysis from video input without retraining or manual intervention. Unlike conventional methods that require manual behavior annotation, video segmentation, task-specific model training, BehaveAgent leverages the reasoning capabilities of multimodal large language models (LLM) to generalize across novel behavioral domains without need for additional training. It integrates LLMs, vision-language models (VLMs), and large-scale visual grounding modules, orchestrated through a multimodal context memory and goal-directed attention mechanism, to enable robust zero-shot visual reasoning across species and experimental paradigms, including plants, insects, rodents, primates, and humans. Upon receiving a video input, BehaveAgent autonomously identifies the correct analysis strategy and performs end-to-end behavior analysis and interpretation without human supervision. Leveraging vision-language representations, it performs general-purpose tracking, pose estimation and segmentation. We demonstrate BehaveAgent’s universal applicability to autonomously (1) identify the behavioral paradigm and develop an action plan specialized for the identified paradigm, (2) identify relevant subjects and objects, (3) track those features, (4) identify behavioral sequences with explicit reasoning, (5) generate and execute code for targeted analysis and (6) generate comprehensive research reports that integrate behavioral findings with relevant scientific literature. Through interpretable agentic reasoning, BehaveAgent makes its internal decision-making process transparent, clarifying why particular features are tracked or behaviors inferred. By reducing the time and expertise required for behavior analysis, BehaveAgent introduces a scalable, generalizable, and explainable paradigm for advancing biological and behavioral research.

Recently, Large Language Model based Autonomous System (LLMAS) has gained great popularity for its potential to simulate complicated behaviors of human societies. One of its main challenges is to present and analyze the dynamic events evolution of LLMAS. In this work, we present a visualization approach to explore the detailed statuses and agents’ behavior within LLMAS. Our approach outlines a general pipeline that organizes raw execution events from LLMAS into a structured behavior model. We leverage a behavior summarization algorithm to create a hierarchical summary of these behaviors, arranged according to their sequence over time. Additionally, we design a cause trace method to mine the causal relationship between agent behaviors. We then develop AgentLens, a visual analysis system that leverages a hierarchical temporal visualization for illustrating the evolution of LLMAS, and supports users to interactively investigate details and causes of agents’ behaviors. Two usage scenarios and a user study demonstrate the effectiveness and usability of our AgentLens.

The trajectory prediction module of autonomous driving navigation systems is required to correctly forecast the future trajectories of other agents and make correct driving decisions to guarantee the safety of self-driving vehicles. However, the trajectory prediction module is susceptible to adversarial attacks, which can lead to wrong driving decisions. Here, we propose RVTP, a robust prediction method against adversarial attacks for trajectory prediction, which is grounded in causal inference. RVTP employs counterfactual intervention to remove the influence of adversarial perturbations on observed history trajectories. Firstly, we present four evaluation metrics to measure directed attacks from four directions under adversarial attacks. Secondly, we establish the causal graph in attack scenarios and study the causal relationships of various elements under attack. Thirdly, we implement the counterfactual intervention based on the causal graph to calculate the causal effect for mitigating the influence of adversarial attacks. Compared with other trajectory prediction methods under attack, extensive experiments demonstrate that RVTP achieves enhanced performance under attacks at the cost of a minimal performance decrease in no attack scenario.

Explanations for AI models in high-stakes domains like medicine often lack verifiability, which can hinder trust. To address this, we propose an interactive agent that produces explanations through an auditable sequence of actions. The agent learns a policy to strategically seek external visual evidence to support its diagnostic reasoning. This policy is optimized using reinforcement learning, resulting in a model that is both efficient and generalizable. Our experiments show that this action-based reasoning process significantly improves calibrated accuracy, reducing the Brier score by 18\% compared to a non-interactive baseline. To validate the faithfulness of the agent's explanations, we introduce a causal intervention method. By masking the visual evidence the agent chooses to use, we observe a measurable degradation in its performance ($\Delta$Brier=+0.029), confirming that the evidence is integral to its decision-making process. Our work provides a practical framework for building AI systems with verifiable and faithful reasoning capabilities.

No abstract available

Trajectory prediction models in autonomous driving are vulnerable to perturbations from non-causal agents whose actions should not affect the ego-agent's behavior. Such perturbations can lead to incorrect predictions of other agents' trajectories, potentially compromising the safety and efficiency of the ego-vehicle's decision-making process. Motivated by this challenge, we propose Causal tRajecTory predICtion (CRiTIC), a novel model that utilizes a causal discovery network to identify inter-agent causal relations over a window of past time steps. To incorporate discovered causal relationships, we propose a novel Causal Attention Gating mechanism to selectively filter information in the proposed Transformer- based architecture. We conduct extensive experiments on two autonomous driving benchmark datasets to evaluate the robustness of our model against non-causal perturbations and its generalization capacity. Our results indicate that the robustness of predictions can be improved by up to 54% without a significant detriment to prediction accuracy. Lastly, we demonstrate the superior domain generalizability of the proposed model, which achieves up to 29% improvement in cross-domain performance. These results underscore the potential of our model to enhance both robustness and generalization capacity for trajectory prediction in diverse autonomous driving domairis.4

Intelligent agents equipped with causal knowledge can optimize their action spaces to avoid unnecessary exploration. The structural causal bandit framework provides a graphical characterization for identifying actions that are unable to maximize rewards by leveraging prior knowledge of the underlying causal structure. While such knowledge enables an agent to estimate the expected rewards of certain actions based on others in online interactions, there has been little guidance on how to transfer information inferred from arbitrary combinations of datasets collected under different conditions -- observational or experimental -- and from heterogeneous environments. In this paper, we investigate the structural causal bandit with transportability, where priors from the source environments are fused to enhance learning in the deployment setting. We demonstrate that it is possible to exploit invariances across environments to consistently improve learning. The resulting bandit algorithm achieves a sub-linear regret bound with an explicit dependence on informativeness of prior data, and it may outperform standard bandit approaches that rely solely on online learning.

To address the inefficiency, resource redundancy, and interpretability challenges in complex agent systems, this paper proposes the Dynamic Hot Path Construction Framework (AHPF). By identifying and optimizing critical high-frequency action sequences (hot paths) through a threedimensional evaluation model (execution frequency, task contribution, path entropy) and causal verification, AHPF integrates multi-modal observation, improved algorithms (reducing 90% redundancy), and path compression $(58 \%$ parameter reduction). Experiments across navigation, dialogue, and microservice scenarios show 32.7% average task efficiency improvement, 56% resource savings, and $\gt79 \%$ hot path coverage. The framework reduces noncritical loads (e.g., 43% energy savings), enhances transparency via visualization, and supports lightweight deployment. Future work focuses on real-time adaptation and safety compliance for broader applications in smart systems.

The large language model (LLM) has achieved significant success across various domains. However, the inherent complexity of causal problems and causal theory poses challenges in accurately describing them in natural language, making it difficult for LLM to comprehend and use them effectively. Causal methods are not easily conveyed through natural language, which hinders LLM's ability to apply them accurately. Additionally, causal datasets are typically tabular, while LLM excels in handling natural language data, creating a structural mismatch that impedes effective reasoning with tabular data. To address these challenges, we have equipped the LLM with causal tools within an agent framework, named the Causal Agent, enabling it to tackle causal problems. The causal agent comprises tools, memory, and reasoning modules. In the tool module, the causal agent calls Python code and uses the encapsulated causal function module to align tabular data with natural language. In the reasoning module, the causal agent performs reasoning through multiple iterations with the tools. In the memory module, the causal agent maintains a dictionary instance where the keys are unique names and the values are causal graphs. To verify the causal ability of the causal agent, we established a Causal Tabular Question Answer (CausalTQA) benchmark consisting of four levels of causal problems: variable level, edge level, causal graph level, and causal effect level. CausalTQA consists of about 1.4K for these four levels questions. Causal agent demonstrates remarkable efficacy on the four-level causal problems, with accuracy rates all above 80\%. Through verification on the real-world dataset QRData, the causal agent is 6\% higher than the original SOTA. For further insights and implementation details, our code is accessible via the GitHub repository https://github.com/kairong-han/causal_agent.

In dynamic home scenarios, due to the complexity of user behavior, the dynamics of environmental conditions, and the fuzziness of multimodal information, the accuracy of intention judgment by service robots has always been relatively low. The degree of association in traditional models depends on statistical correlation, which leads to inaccurate causal judgments and severe interaction failures. Therefore, this study proposes a multimodal causal reasoning model based on the characteristics of dynamic family scenarios, quantitatively determining the corresponding relationship between the key feature information of dynamic family scenarios and the failure transmission mechanism of intention judgment, and constructing a coupling model of influencing factors and interaction failure. Based on the establishment of a causal reasoning model integrating multimodal information such as speech, vision and touch, the intention is modified by using the structural causal model (SCM) and counterfactual causal reasoning, and a causal confidence decision-making framework is designed to achieve early prediction and hierarchical avoidance of interaction failure. The results show that for the above-mentioned real-life interaction scenarios, the intent recognition accuracy of this framework in dynamic scenarios is 24.3% higher than that of the framework using only Transformer fusion, the recognition error rate is reduced by 62.1%, and the number of invalid interactions is decreased by 67%. Meanwhile, it can still maintain an accuracy of over 78% under sudden interferences (such as pet interference, light changes, etc.). This research provides a theoretical reference for the stable decision-making of service robots in a highly dynamic environment.

Human conflict is often attributed to threats against material conditions and symbolic values, yet it remains unclear how they interact and which dominates. Progress is limited by weak causal control, ethical constraints, and scarce temporal data. We address these barriers using simulations of large language model (LLM)-driven agents in virtual societies, independently varying realistic and symbolic threat while tracking actions, language, and attitudes. Representational analyses show that the underlying LLM encodes realistic threat, symbolic threat, and hostility as distinct internal states, that our manipulations map onto them, and that steering these states causally shifts behavior. Our simulations provide a causal account of threat-driven conflict over time: realistic threat directly increases hostility, whereas symbolic threat effects are weaker, fully mediated by ingroup bias, and increase hostility only when realistic threat is absent. Non-hostile intergroup contact buffers escalation, and structural asymmetries concentrate hostility among majority groups.

Automatic Question Generation (AQG) aims to generate valid, coherent questions based on given text passages in pre-trained language models. While AQG has been a significant area of retrieval augmentation and agent-based systems, current QG models face limitations, especially with sequential models like Transformer, which struggle with modeling complex logical structures and limited question coherence and depth. This paper proposes Q-Chain, a framework designed to optimize logic and educational values. Q-Chain features: (1)Differentiable logic layers in GMP model conditional dependencies and counterfactuals, ensuring logical rigor. (2) Bloom's taxonomy-guided gating adjusts question difficulty to align with pedagogical goals. (3) Direct generation of logic-structured question graphs enhanced by counterfactual training. Experimental results show Q-Chain outperforms GPT-3.5, T5, and BART on the MedQuAD dataset (0.82 vs 0.80 vs 0.72 and 0.74 F1-score) and shows superior robustness to noisy inputs, achieving a 4.1/5 human rating on counterfactual questions, 40% better than BART.

Real-world decision-making problems are often marked by complex, uncertain dynamics that can shift or break under changing conditions. Traditional Model-Based Reinforcement Learning (MBRL) approaches learn predictive models of environment dynamics from queried trajectories and then use these models to simulate rollouts for policy optimization. However, such methods do not account for the underlying causal mechanisms that govern the environment, and thus inadvertently capture spurious correlations, making them sensitive to distributional shifts and limiting their ability to generalize. The same naturally holds for model-free approaches. In this work, we introduce Causal Model-Based Policy Optimization (C-MBPO), a novel framework that integrates causal learning into the MBRL pipeline to achieve more robust, explainable, and generalizable policy learning algorithms. Our approach centers on first inferring a Causal Markov Decision Process (C-MDP) by learning a local Structural Causal Model (SCM) of both the state and reward transition dynamics from trajectories gathered online. C-MDPs differ from classic MDPs in that we can decompose causal dependencies in the environment dynamics via specifying an associated Causal Bayesian Network. C-MDPs allow for targeted interventions and counterfactual reasoning, enabling the agent to distinguish between mere statistical correlations and causal relationships. The learned SCM is then used to simulate counterfactual on-policy transitions and rewards under hypothetical actions (or ``interventions"), thereby guiding policy optimization more effectively. The resulting policy learned by C-MBPO can be shown to be robust to a class of distributional shifts that affect spurious, non-causal relationships in the dynamics. We demonstrate this through some simple experiments involving near and far OOD dynamics drifts.

As Multi Agent Reinforcement Learning systems are used in safety critical applications. Understanding why agents make decisions and how they achieve collective behavior is crucial. Existing explainable AI methods struggle in multi agent settings. They fail to attribute collective outcomes to individuals, quantify emergent behaviors, or capture complex interactions. We present MACIE Multi Agent Causal Intelligence Explainer, a framework combining structural causal models, interventional counterfactuals, and Shapley values to provide comprehensive explanations. MACIE addresses three questions. First, each agent's causal contribution using interventional attribution scores. Second, system level emergent intelligence through synergy metrics separating collective effects from individual contributions. Third, actionable explanations using natural language narratives synthesizing causal insights. We evaluate MACIE across four MARL scenarios: cooperative, competitive, and mixed motive. Results show accurate outcome attribution, mean phi_i equals 5.07, standard deviation less than 0.05, detection of positive emergence in cooperative tasks, synergy index up to 0.461, and efficient computation, 0.79 seconds per dataset on CPU. MACIE uniquely combines causal rigor, emergence quantification, and multi agent support while remaining practical for real time use. This represents a step toward interpretable, trustworthy, and accountable multi agent AI.

Exploration in sparse-reward tasks remains a fundamental challenge in multi-agent reinforcement learning (MARL) due to complex inter-agent interactions and the expansive exploration space. To address this issue, we propose Targeted Multi-Agent Exploration (TMAE), a novel framework that uncovers the causal relationships between the state space and the reward function, thereby reducing the exploration space and enabling more targeted exploration. Specifically, we construct a structural causal model (SCM) to model the causality between sub-state variables and sparse rewards, providing a robust analytical foundation for subsequent causal inference. Through counterfactual causal intervention, TMAE identifies the most critical subspaces for discovering rare but pivotal events while filtering out confounders. By incorporating these causal insights into the exploration process, TMAE prioritizes subspaces with stronger causal effects on sparse rewards, significantly enhancing exploration efficiency. We evaluate TMAE on a range of MARL benchmarks featuring sparse rewards, consistently demonstrating superior exploration efficiency compared to state-of-the-art methods. Furthermore, visualized causal insights derived from TMAE reveal its ability to effectively capture intricate dependencies and priorities in targeted exploration, showcasing strong alignment with prior domain knowledge.

This paper addresses the prevalent pain points in the equipment health management systems of nuclear power units, including difficulties in identifying model alarm deviations (such as missed alarms and false alarms), complex attribution analysis, and low optimization efficiency. It proposes and develops an alarm deviation optimization agent based on a large language model (LLM).With human-machine collaboration as its core concept, this agent integrates cutting-edge technologies including retrieval-augmented generation (RAG), database question-answering, external tool calling, and agent cognitive architecture, and constructs a full-process automated solution covering “deviation detection—root cause analysis—optimization recommendation—closed-loop management.”By interfacing with the electronic log database and equipment health management platform of nuclear power plants, the agent automatically compares manual work orders with system alarm information to accurately identify the two types of deviations. On this basis, it innovatively introduces the ReAct reasoning paradigm and counterfactual scenario reconstruction mechanism, and combines with the structural causal model (SCM) to conduct in-depth attribution analysis on multi-dimensional factors such as data pipelines, model logic, and threshold settings. Meanwhile, it leverages RAG technology to enhance the large model's semantic understanding of complex model configurations. Additionally, a multi-source evidence fusion framework is built to achieve self-consistency verification through four dimensions of evidence (numerical, event-based, contextual, and analogical), and the Reflextion mechanism is introduced to support the agent's self-reflection and continuous optimization.Practical applications show that the agent has significantly improved the operational efficiency and accuracy of the platform, providing a replicable technical paradigm for the intelligent upgrade of industrial equipment health management.

In human-AI interaction, a prominent goal is to attain human`s desirable outcome with the assistance of AI agents, which can be ideally delineated as a problem of seeking the optimal Nash Equilibrium that matches the human`s desirable outcome. However, reaching the outcome is usually challenging due to the existence of multiple Nash Equilibria that are related to the assisting task but do not correspond to the human`s desirable outcome. To tackle this issue, we employ a theoretical framework called structural causal game (SCG) to formalize the human-AI interactive process. Furthermore, we introduce a strategy referred to as pre-policy intervention on the SCG to steer AI agents towards attaining the human`s desirable outcome. In more detail, a pre-policy is learned as a generalized intervention to guide the agents` policy selection, under a transparent and interpretable procedure determined by the SCG. To make the framework practical, we propose a reinforcement learning-like algorithm to search out this pre-policy. The proposed algorithm is tested in both gridworld environments and realistic dialogue scenarios with large language models, demonstrating its adaptability in a broader class of problems and potential effectiveness in real-world situations.

Stochastic sequential decision-making systems — such as Markov decision processes and their variants — are increasingly used in areas such as transportation, healthcare, and communication. However, the ability to explain these systems’ outputs to non-technical end users has not kept pace with their widespread adoption. This paper addresses that gap by extending prior work and presenting a unified framework for generating causal explanations of agent behavior in sequential decision-making settings, grounded in the structural causal model (SCM) paradigm. Our framework supports the generation of multiple, semantically distinct explanations for agent actions — capabilities that were previously unattainable. In addition to introducing a novel taxonomy of explanations for MDPs to guide empirical investigation, we develop both exact and approximate causal inference methods within the SCM framework. We analyze their applicability and derive run-time bounds for each. This leads to the proposed algorithm, MeanRESP, which operates flexibly across a spectrum of approximations tailored to external constraints. We further analyze the sample complexity and error rates of approximate MeanRESP, and provide a detailed comparison of its outputs—under varying definitions of responsibility—with popular Shapley-value-based methods. Empirically, we performed a series of experiments to evaluate the practicality and effectiveness of the proposed system, focusing on real-world computational demands and the validity and reliability of metrics for comparing approximate and exact causal methods. Finally, we present two user studies that reveal user preferences for certain types of explanations and demonstrate a strong preference for explanations generated by our framework compared to those from other state-of-the-art systems.

Loneliness is a prevalent and global problem for all people, and it can adversely affect their quality of life. Many research investigations have confirmed the negative psychological impacts of loneliness on people’s the unwanted impact of loneliness. Yet, these interventions are missing the power of reasoning to predict the onset of loneliness. Consequently, this paper presented the work of developing a computational cognitive agent model of loneliness using a causal networking modeling approach by relying on discrepancy model as a benchmark to serve as analytics engine for a companion robot design. Loneliness determinants and their causal relationships were identified from the literature and formalized to construct the intended cognitive agent model. Furthermore, simulation analyses under various parameter settings were implemented to explore the causal relationships among the identified loneliness determinants and those simulations revealed similar behaviors or patterns to existing literature. The designed cognitive agent model was evaluated using both of mathematical analysis and automated logical analysis. These two evaluation approaches have proved the correctness of the designed model. The developed computational loneliness agent model with little tuning can serve as a core analytical engine for intelligent technologies such as robots to control and monitor the adverse effects of loneliness

Multi-Agent Path Finding refers to the problem of finding the optimal path set for multiple agents from the starting position to the target position without conflict. In multi-agent path planning, multiple factors need to be considered, such as the dynamic characteristics of the agent, environmental constraints, communication and collaboration mechanisms, etc. In order to solve such problems, researchers have proposed a variety of path planning algorithms, including centralized path planning algorithms, distributed path planning algorithms and collaborative path planning algorithms. In this paper, we propose a decentralized multi-agent path planning framework that combines causal relationship drive and graph neural network(CRGN). First, we propose an end-to-end framework to extract global features and local features in images. On this basis, we design the fusion of deep features and shallow features to increase visual fine-graining. Then, we propose to use graph neural networks to aggregate information communication between agents and build a dynamic communication topology network through agent communication within the local field of view. Finally, we propose a new reinforcement learning algorithm to train the agent’s generalization ability in unknown environments. We fit the state value function and action value function in reinforcement learning through deep neural networks. Thus, it can better solve the intelligent agent path planning under large-scale complex problems. We verified the effectiveness of the model on multiple different datasets. Experimental results show that our proposed approach is close to the performance of expert algorithms. At the same time, we show through extensive ablation experiments that our model can exhibit good performance and robustness even in unseen scenes.

Structural Causal Explanations (SCEs) can be used to automatically generate explanations in natural language to questions about given data that are grounded in a (possibly learned) causal model. Unfortunately they work for small data only. In turn they are not attractive to offer reasons for events, e.g., tracking causal changes over multiple time steps, or a behavioral component that involves feedback loops through actions of an agent. To this end, we generalize SCEs to a (recursive) formulation of explanation trees to capture the temporal interactions between reasons. We show the benefits of this more general SCE algorithm on synthetic time-series data and a 2D grid game, and further compare it to the base SCE and other existing methods for causal explanations.

We present a novel autonomous driving framework, DualAD, designed to imitate human reasoning during driving. DualAD comprises two layers: a rule-based motion planner at the bottom layer that handles routine driving tasks requiring minimal reasoning, and an upper layer featuring a rule-based text encoder that converts driving scenarios from absolute states into text description. This text is then processed by a large language model (LLM) to make driving decisions. The upper layer intervenes in the bottom layer's decisions when potential danger is detected, mimicking human reasoning in critical situations. Closed-loop experiments demonstrate that DualAD, using a zero-shot pre-trained model, significantly outperforms rule-based motion planners that lack reasoning abilities. Our experiments also highlight the effectiveness of the text encoder, which considerably enhances the model's scenario understanding. Additionally, the integrated DualAD model improves with stronger LLMs, indicating the framework's potential for further enhancement. Code and benchmarks are available at github.com/TUM-AVS/DualAD.

Reasoning is a fundamental cognitive process underlying inference, problem-solving, and decision-making. While large language models (LLMs) demonstrate strong reasoning capabilities in closed-world settings, they struggle in open-ended and dynamic environments. Agentic reasoning marks a paradigm shift by reframing LLMs as autonomous agents that plan, act, and learn through continual interaction. In this survey, we organize agentic reasoning along three complementary dimensions. First, we characterize environmental dynamics through three layers: foundational agentic reasoning, which establishes core single-agent capabilities including planning, tool use, and search in stable environments; self-evolving agentic reasoning, which studies how agents refine these capabilities through feedback, memory, and adaptation; and collective multi-agent reasoning, which extends intelligence to collaborative settings involving coordination, knowledge sharing, and shared goals. Across these layers, we distinguish in-context reasoning, which scales test-time interaction through structured orchestration, from post-training reasoning, which optimizes behaviors via reinforcement learning and supervised fine-tuning. We further review representative agentic reasoning frameworks across real-world applications and benchmarks, including science, robotics, healthcare, autonomous research, and mathematics. This survey synthesizes agentic reasoning methods into a unified roadmap bridging thought and action, and outlines open challenges and future directions, including personalization, long-horizon interaction, world modeling, scalable multi-agent training, and governance for real-world deployment.

Text-to-image generation has advanced rapidly with large-scale multimodal training, yet fine-grained controllability remains a critical challenge. Counterfactual controllability, defined as the capacity to deliberately generate images that contradict common-sense patterns, remains a major challenge but plays a crucial role in enabling creativity and exploratory applications. In this work, we address this gap with a focus on counterfactual size (e.g., generating a tiny walrus beside a giant button) and propose an automatic prompt engineering framework that adapts base prompts into revised prompts for counterfactual images. The framework comprises three components: an image evaluator that guides dataset construction by identifying successful image generations, a supervised prompt rewriter that produces revised prompts, and a DPO-trained ranker that selects the optimal revised prompt. We construct the first counterfactual size text-image dataset and enhance the image evaluator by extending Grounded SAM with refinements, achieving a 114 percent improvement over its backbone. Experiments demonstrate that our method outperforms state-of-the-art baselines and ChatGPT-4o, establishing a foundation for future research on counterfactual controllability.

Recent Large Audio Language Models (LALMs) excel in understanding but often lack transparent reasoning. To address this "black-box" limitation, we organized the Audio Reasoning Challenge at Interspeech 2026, the first shared task dedicated to evaluating Chain-of-Thought (CoT) quality in the audio domain. The challenge introduced MMAR-Rubrics, a novel instance-level protocol assessing the factuality and logic of reasoning chains. Featured Single Model and Agent tracks, the competition attracting 156 teams from 18 countries and regions. Results show agent systems currently lead in reasoning quality, utilizing iterative tool orchestration and cross-modal analysis. Besides, single models are rapidly advancing via reinforcement learning and sophisticated data pipeline. We details the challenge design, methodology, and a comprehensive analysis of state-of-the-art systems, providing new insights for explainable audio intelligence.

This paper presents a way of enhancing the reliability of Large Multi-modal Models (LMMs) in addressing hallucination, where the models generate cross-modal inconsistent responses. Without additional training, we propose Counterfactual Inception, a novel method that implants counterfactual thinking into LMMs using self-generated counterfactual keywords. Our method is grounded in the concept of counterfactual thinking, a cognitive process where human considers alternative realities, enabling more extensive context exploration. Bridging the human cognition mechanism into LMMs, we aim for the models to engage with and generate responses that span a wider contextual scene understanding, mitigating hallucinatory outputs. We further introduce Plausibility Verification Process (PVP), a simple yet robust keyword constraint that effectively filters out sub-optimal keywords to enable the consistent triggering of counterfactual thinking in the model responses. Comprehensive analyses across various LMMs, including both open-source and proprietary models, corroborate that counterfactual thinking significantly reduces hallucination and helps to broaden contextual understanding based on true visual clues.

Large language models (LLMs) have shown strong abilities in reasoning and problem solving, but recent studies reveal that they still struggle with time series reasoning tasks, where outputs are often affected by hallucination or knowledge leakage. In this work we propose TS-Agent, a time series reasoning agent that leverages LLMs strictly for what they excel at, i.e., gathering evidence and synthesizing it into conclusions through step-by-step reasoning, while delegating the extraction of statistical and structural information to time series analytical tools. Instead of mapping time series into text tokens, images, or embeddings, our agent interacts with raw numeric sequences through atomic operators, records outputs in an explicit evidence log, and iteratively refines its reasoning under the guidance of a self-critic and a final quality gate. This design avoids multi-modal alignment training, preserves the native form of time series, ensures interpretability and verifiability, and mitigates knowledge leakage or hallucination. Empirically, we evaluate the agent on established benchmarks. Our experiments show that TS-Agent achieves performance comparable to state-of-the-art LLMs on understanding benchmarks, and delivers significant improvements on reasoning tasks, where existing models often rely on memorization and fail in zero-shot settings.

Large language models (LLMs) excel at natural language tasks but remain brittle in domains requiring precise logical and symbolic reasoning. Chaotic dynamical systems provide an especially demanding test because chaos is deterministic yet often misinterpreted as randomness or complexity. We introduce ChaosBench-Logic, a benchmark that evaluates LLM reasoning across 30 diverse dynamical systems using a unified first-order logic (FOL) ontology. Each system is annotated with truth assignments for 11 semantic predicates, and 621 questions are generated across seven reasoning categories, including multi-hop implications, cross-system analogies, counterfactual reasoning, bias probes, and multi-turn dialogues. We define metrics for logical accuracy, implication consistency, dialogue coherence, and contradiction, and we release an open-source evaluation pipeline. Initial experiments show that frontier LLMs such as GPT-4, Claude 3.5 Sonnet, Gemini 2.5 Flash, and the open-source LLaMA-3 70B achieve 91-94% per-item accuracy, yet still score 0% on compositional items and exhibit fragile global coherence. Dialogue-level accuracy ranges from 53.1% (GPT-4 CoT) to 75.5% (LLaMA-3 zero-shot). ChaosBench-Logic provides a rigorous testbed for diagnosing such failures and a foundation for developing neuro-symbolic approaches that improve scientific reasoning in LLMs.

Language agents perform complex tasks by using tools to execute each step precisely. However, most existing agents are based on proprietary models or designed to target specific tasks, such as mathematics or multi-hop question answering. We introduce Husky, a holistic, open-source language agent that learns to reason over a unified action space to address a diverse set of complex tasks involving numerical, tabular, and knowledge-based reasoning. Husky iterates between two stages: 1) generating the next action to take towards solving a given task and 2) executing the action using expert models and updating the current solution state. We identify a thorough ontology of actions for addressing complex tasks and curate high-quality data to train expert models for executing these actions. Our experiments show that Husky outperforms prior language agents across 14 evaluation datasets. Moreover, we introduce HuskyQA, a new evaluation set which stress tests language agents for mixed-tool reasoning, with a focus on retrieving missing knowledge and performing numerical reasoning. Despite using 7B models, Husky matches or even exceeds frontier LMs such as GPT-4 on these tasks, showcasing the efficacy of our holistic approach in addressing complex reasoning problems. Our code and models are available at https://github.com/agent-husky/Husky-v1.

We present STAgent, an agentic large language model tailored for spatio-temporal understanding, designed to solve complex tasks such as constrained point-of-interest discovery and itinerary planning. STAgent is a specialized model capable of interacting with ten distinct tools within spatio-temporal scenarios, enabling it to explore, verify, and refine intermediate steps during complex reasoning. Notably, STAgent effectively preserves its general capabilities. We empower STAgent with these capabilities through three key contributions: (1) a stable tool environment that supports over ten domain-specific tools, enabling asynchronous rollout and training; (2) a hierarchical data curation framework that identifies high-quality data like a needle in a haystack, curating high-quality queries by retaining less than 1\% of the raw data, emphasizing both diversity and difficulty; and (3) a cascaded training recipe that starts with a seed SFT stage acting as a guardian to measure query difficulty, followed by a second SFT stage fine-tuned on queries with high certainty, and an ultimate RL stage that leverages data of low certainty. Initialized with Qwen3-30B-A3B to establish a strong SFT foundation and leverage insights into sample difficulty, STAgent yields promising performance on TravelBench while maintaining its general capabilities across a wide range of general benchmarks, thereby demonstrating the effectiveness of our proposed agentic model.

AI-augmented systems are traditionally designed to streamline human decision-making by minimizing cognitive load, clarifying arguments, and optimizing efficiency. However, in a world where algorithmic certainty risks becoming an Orwellian tool of epistemic control, true intellectual growth demands not passive acceptance but active struggle. Drawing on the dystopian visions of George Orwell and Philip K. Dick - where reality is unstable, perception malleable, and truth contested - this paper introduces Cognitive Dissonance AI (CD-AI): a novel framework that deliberately sustains uncertainty rather than resolving it. CD-AI does not offer closure, but compels users to navigate contradictions, challenge biases, and wrestle with competing truths. By delaying resolution and promoting dialectical engagement, CD-AI enhances reflective reasoning, epistemic humility, critical thinking, and adaptability in complex decision-making. This paper examines the theoretical foundations of the approach, presents an implementation model, explores its application in domains such as ethics, law, politics, and science, and addresses key ethical concerns - including decision paralysis, erosion of user autonomy, cognitive manipulation, and bias in AI reasoning. In reimagining AI as an engine of doubt rather than a deliverer of certainty, CD-AI challenges dominant paradigms of AI-augmented reasoning and offers a new vision - one in which AI sharpens the mind not by resolving conflict, but by sustaining it. Rather than reinforcing Huxleyan complacency or pacifying the user into intellectual conformity, CD-AI echoes Nietzsche's vision of the Uebermensch - urging users to transcend passive cognition through active epistemic struggle.

In causal inference, whether through randomized controlled trials or observational studies, access to both treated and control units is essential for estimating the effect of a treatment on an outcome of interest. When treatment assignment is random, the average treatment effect (ATE) can be estimated directly by comparing outcomes between groups. In non-randomized settings, various techniques are employed to adjust for confounding and approximate the counterfactual scenario to recover an unbiased ATE. A common challenge, especially in observational studies, is the absence of units clearly labeled as controls-that is, units known not to have received the treatment. To address this, we propose positive-unlabeled (PU) learning as a framework for identifying, with high confidence, control units from a pool of unlabeled ones, using only the available treated (positive) units. We evaluate this approach using both simulated and real-world data. We construct a causal graph with diverse relationships and use it to generate synthetic data under various scenarios, assessing how reliably the method recovers control groups that allow estimates of true ATE. We also apply our approach to real-world data on optimal sowing and fertilizer treatments in sustainable agriculture. Our findings show that PU learning can successfully identify control (negative) units from unlabeled data based only on treated units and, through the resulting control group, estimate an ATE that closely approximates the true value. This work has important implications for observational causal inference, especially in fields where randomized experiments are difficult or costly. In domains such as earth, environmental, and agricultural sciences, it enables a plethora of quasi-experiments by leveraging available earth observation and climate data, particularly when treated units are available but control units are lacking.

Various methods have recently been proposed to estimate causal effects with confidence intervals that are uniformly valid over a set of data generating processes when high-dimensional nuisance models are estimated by post-model-selection or machine learning estimators. These methods typically require that all the confounders are observed to ensure identification of the effects. We contribute by showing how valid semiparametric inference can be obtained in the presence of unobserved confounders and high-dimensional nuisance models. We propose uncertainty intervals which allow for unobserved confounding, and show that the resulting inference is valid when the amount of unobserved confounding is small relative to the sample size; the latter is formalized in terms of convergence rates. Simulation experiments illustrate the finite sample properties of the proposed intervals and investigate an alternative procedure that improves the empirical coverage of the intervals when the amount of unobserved confounding is large. Finally, a case study on the effect of smoking during pregnancy on birth weight is used to illustrate the use of the methods introduced to perform a sensitivity analysis to unobserved confounding.

Lifted inference exploits symmetries in probabilistic graphical models by using a representative for indistinguishable objects, thereby speeding up query answering while maintaining exact answers. Even though lifting is a well-established technique for the task of probabilistic inference in relational domains, it has not yet been applied to the task of causal inference. In this paper, we show how lifting can be applied to efficiently compute causal effects in relational domains. More specifically, we introduce parametric causal factor graphs as an extension of parametric factor graphs incorporating causal knowledge and give a formal semantics of interventions therein. We further present the lifted causal inference algorithm to compute causal effects on a lifted level, thereby drastically speeding up causal inference compared to propositional inference, e.g., in causal Bayesian networks. In our empirical evaluation, we demonstrate the effectiveness of our approach.

Practitioners making decisions based on causal effects typically ignore structural uncertainty. We analyze when this uncertainty is consequential enough to warrant methodological solutions (Bayesian model averaging over competing causal structures). Focusing on bivariate relationships ($X \rightarrow Y$ vs. $X \leftarrow Y$), we establish that model averaging is beneficial when: (1) structural uncertainty is moderate to high, (2) causal effects differ substantially between structures, and (3) loss functions are sufficiently sensitive to the size of the causal effect. We prove optimality results of our suggested methodological solution under regularity conditions and demonstrate through simulations that modern causal discovery methods can provide, within limits, the necessary quantification. Our framework complements existing robust causal inference approaches by addressing a distinct source of uncertainty typically overlooked in practice.

Knowledge of the underlying causal relations is essential for inferring the effect of interventions in complex systems. In a widely studied approach, structural causal models postulate noisy functional relations among interacting variables, where the underlying causal structure is then naturally represented by a directed graph whose edges indicate direct causal dependencies. In the typical application, this underlying causal structure must be learned from data, and thus, the remaining structure uncertainty needs to be incorporated into causal inference in order to draw reliable conclusions. In recent work, test inversions provide an ansatz to account for this data-driven model choice and, therefore, combine structure learning with causal inference. In this article, we propose the use of dual likelihood to greatly simplify the treatment of the involved testing problem. Indeed, dual likelihood leads to a closed-form solution for constructing confidence regions for total causal effects that rigorously capture both sources of uncertainty: causal structure and numerical size of nonzero effects. The proposed confidence regions can be computed with a bottom-up procedure starting from sink nodes. To render the causal structure identifiable, we develop our ideas in the context of linear causal relations with equal error variances.

White-box AI (WAI), or explainable AI (XAI) model, a novel tool to achieve the reasoning behind decisions and predictions made by the AI algorithms, makes it more understandable and transparent. It offers a new approach to address key challenges of interpretability and mathematical validation in traditional black-box models. In this paper, WAI-aided wireless communication systems are proposed and investigated thoroughly to utilize the promising capabilities. First, we introduce the fundamental principles of WAI. Then, a detailed comparison between WAI and traditional black-box model is conducted in terms of optimization objectives and architecture design, with a focus on deep neural networks (DNNs) and transformer networks. Furthermore, in contrast to the traditional black-box methods, WAI leverages theory-driven causal modeling and verifiable optimization paths, thereby demonstrating potential advantages in areas such as signal processing and resource allocation. Finally, we outline future research directions for the integration of WAI in wireless communication systems.

Counterfactual decision-making in the face of uncertainty involves selecting the optimal action from several alternatives using causal reasoning. Decision-makers often rank expected potential outcomes (or their corresponding utility and desirability) to compare the preferences of candidate actions. In this paper, we study new counterfactual decision-making rules by introducing two new metrics: the probabilities of potential outcome ranking (PoR) and the probability of achieving the best potential outcome (PoB). PoR reveals the most probable ranking of potential outcomes for an individual, and PoB indicates the action most likely to yield the top-ranked outcome for an individual. We then establish identification theorems and derive bounds for these metrics, and present estimation methods. Finally, we perform numerical experiments to illustrate the finite-sample properties of the estimators and demonstrate their application to a real-world dataset.

This paper introduces SOLID (Synergizing Optimization and Large Language Models for Intelligent Decision-Making), a novel framework that integrates mathematical optimization with the contextual capabilities of large language models (LLMs). SOLID facilitates iterative collaboration between optimization and LLMs agents through dual prices and deviation penalties. This interaction improves the quality of the decisions while maintaining modularity and data privacy. The framework retains theoretical convergence guarantees under convexity assumptions, providing insight into the design of LLMs prompt. To evaluate SOLID, we applied it to a stock portfolio investment case with historical prices and financial news as inputs. Empirical results demonstrate convergence under various scenarios and indicate improved annualized returns compared to a baseline optimizer-only method, validating the synergy of the two agents. SOLID offers a promising framework for advancing automated and intelligent decision-making across diverse domains.

The growing integration of robots in shared environments - such as warehouses, shopping centres, and hospitals - demands a deep understanding of the underlying dynamics and human behaviours, including how, when, and where individuals engage in various activities and interactions. This knowledge goes beyond simple correlation studies and requires a more comprehensive causal analysis. By leveraging causal inference to model cause-and-effect relationships, we can better anticipate critical environmental factors and enable autonomous robots to plan and execute tasks more effectively. To this end, we propose a novel causality-based decision-making framework that reasons over a learned causal model to assist the robot in deciding when and how to complete a given task. In the examined use case - i.e., a warehouse shared with people - we exploit the causal model to estimate battery usage and human obstructions as factors influencing the robot's task execution. This reasoning framework supports the robot in making informed decisions about task timing and strategy. To achieve this, we developed also PeopleFlow, a new Gazebo-based simulator designed to model context-sensitive human-robot spatial interactions in shared workspaces. PeopleFlow features realistic human and robot trajectories influenced by contextual factors such as time, environment layout, and robot state, and can simulate a large number of agents. While the simulator is general-purpose, in this paper we focus on a warehouse-like environment as a case study, where we conduct an extensive evaluation benchmarking our causal approach against a non-causal baseline. Our findings demonstrate the efficacy of the proposed solutions, highlighting how causal reasoning enables autonomous robots to operate more efficiently and safely in dynamic environments shared with humans.

Decision support systems based on prediction sets help humans solve multiclass classification tasks by narrowing down the set of potential label values to a subset of them, namely a prediction set, and asking them to always predict label values from the prediction sets. While this type of systems have been proven to be effective at improving the average accuracy of the predictions made by humans, by restricting human agency, they may cause harm$\unicode{x2014}$a human who has succeeded at predicting the ground-truth label of an instance on their own may have failed had they used these systems. In this paper, our goal is to control how frequently a decision support system based on prediction sets may cause harm, by design. To this end, we start by characterizing the above notion of harm using the theoretical framework of structural causal models. Then, we show that, under a natural, albeit unverifiable, monotonicity assumption, we can estimate how frequently a system may cause harm using only predictions made by humans on their own. Further, we also show that, under a weaker monotonicity assumption, which can be verified experimentally, we can bound how frequently a system may cause harm again using only predictions made by humans on their own. Building upon these assumptions, we introduce a computational framework to design decision support systems based on prediction sets that are guaranteed to cause harm less frequently than a user-specified value using conformal risk control. We validate our framework using real human predictions from two different human subject studies and show that, in decision support systems based on prediction sets, there is a trade-off between accuracy and counterfactual harm.

Many researchers have applied classical statistical decision theory to evaluate treatment choices and learn optimal policies. However, because this framework is based solely on realized outcomes under chosen decisions and ignores counterfactual outcomes, it cannot assess the quality of a decision relative to feasible alternatives. For example, in bail decisions, a judge must consider not only crime prevention but also the avoidance of unnecessary burdens on arrestees. To address this limitation, we generalize standard decision theory by incorporating counterfactual losses, allowing decisions to be evaluated using all potential outcomes. The central challenge in this counterfactual statistical decision framework is identification: since only one potential outcome is observed for each unit, the associated counterfactual risk is generally not identifiable. We prove that, under the assumption of strong ignorability, the counterfactual risk is identifiable if and only if the counterfactual loss function is additive in the potential outcomes. Moreover, we demonstrate that additive counterfactual losses can yield treatment recommendations, which differ from those based on standard loss functions when the decision problem involves more than two treatment options. One interpretation of this result is that additive counterfactual losses can capture the accuracy and difficulty of a decision, whereas standard losses account for accuracy alone. Finally, we formulate a symbolic linear inverse program that, given a counterfactual loss, determines whether its risk is identifiable, without requiring data.

Generative models for counterfactual outcomes face two key sources of bias. Confounding bias arises when approaches fail to account for systematic differences between those who receive the intervention and those who do not. Misspecification bias arises when methods attempt to address confounding through estimation of an auxiliary model, but specify it incorrectly. We introduce DoubleGen, a doubly robust framework that modifies generative modeling training objectives to mitigate these biases. The new objectives rely on two auxiliaries -- a propensity and outcome model -- and successfully address confounding bias even if only one of them is correct. We provide finite-sample guarantees for this robustness property. We further establish conditions under which DoubleGen achieves oracle optimality -- matching the convergence rates standard approaches would enjoy if interventional data were available -- and minimax rate optimality. We illustrate DoubleGen with three examples: diffusion models, flow matching, and autoregressive language models.

Counterfactuals are widely used in AI to explain how minimal changes to a model's input can lead to a different output. However, established methods for computing counterfactuals typically focus on one-step decision-making, and are not directly applicable to sequential decision-making tasks. This paper fills this gap by introducing counterfactual strategies for Markov Decision Processes (MDPs). During MDP execution, a strategy decides which of the enabled actions (with known probabilistic effects) to execute next. Given an initial strategy that reaches an undesired outcome with a probability above some limit, we identify minimal changes to the initial strategy to reduce that probability below the limit. We encode such counterfactual strategies as solutions to non-linear optimization problems, and further extend our encoding to synthesize diverse counterfactual strategies. We evaluate our approach on four real-world datasets and demonstrate its practical viability in sophisticated sequential decision-making tasks.

We consider the problem of causal discovery (a.k.a., causal structure learning) in a multi-domain setting. We assume that the causal functions are invariant across the domains, while the distribution of the exogenous noise may vary. Under causal sufficiency (i.e., no confounders exist), we show that the causal diagram can be discovered under less restrictive functional assumptions compared to previous work. What enables causal discovery in this setting is bijective generation mechanisms (BGM), which ensures that the functional relation between the exogenous noise $E$ and the endogenous variable $Y$ is bijective and differentiable in both directions at every level of the cause variable $X = x$. BGM generalizes a variety of models including additive noise model, LiNGAM, post-nonlinear model, and location-scale noise model. Further, we derive a statistical test to find the parents set of the target variable. Experiments on various synthetic and real-world datasets validate our theoretical findings.

Causal inference, a cornerstone in disciplines such as economics, genomics, and medicine, is increasingly being recognized as fundamental to advancing the field of robotics. In particular, the ability to reason about cause and effect from observational data is crucial for robust generalization in robotic systems. However, the construction of a causal graphical model, a mechanism for representing causal relations, presents an immense challenge. Currently, a nuanced grasp of causal inference, coupled with an understanding of causal relationships, must be manually programmed into a causal graphical model. To address this difficulty, we present initial results towards a human-centered augmented reality framework for creating causal graphical models. Concretely, our system bootstraps the causal discovery process by involving humans in selecting variables, establishing relationships, performing interventions, generating counterfactual explanations, and evaluating the resulting causal graph at every step. We highlight the potential of our framework via a physical robot manipulator on a pick-and-place task.

Recent developments enable the quantification of causal control given a structural causal model (SCM). This has been accomplished by introducing quantities which encode changes in the entropy of one variable when intervening on another. These measures, named causal entropy and causal information gain, aim to address limitations in existing information theoretical approaches for machine learning tasks where causality plays a crucial role. They have not yet been properly mathematically studied. Our research contributes to the formal understanding of the notions of causal entropy and causal information gain by establishing and analyzing fundamental properties of these concepts, including bounds and chain rules. Furthermore, we elucidate the relationship between causal entropy and stochastic interventions. We also propose definitions for causal conditional entropy and causal conditional information gain. Overall, this exploration paves the way for enhancing causal machine learning tasks through the study of recently-proposed information theoretic quantities grounded in considerations about causality.

Deploying robots in human-shared spaces requires understanding interactions among nearby agents and objects. Modelling cause-and-effect relations through causal inference aids in predicting human behaviours and anticipating robot interventions. However, a critical challenge arises as existing causal discovery methods currently lack an implementation inside the ROS ecosystem, the standard de facto in robotics, hindering effective utilisation in robotics. To address this gap, this paper introduces ROS-Causal, a ROS-based framework for onboard data collection and causal discovery in human-robot spatial interactions. An ad-hoc simulator, integrated with ROS, illustrates the approach's effectiveness, showcasing the robot onboard generation of causal models during data collection. ROS-Causal is available on GitHub: https://github.com/lcastri/roscausal.git.

Precision rehabilitation offers the promise of an evidence-based approach for optimizing individual rehabilitation to improve long-term functional outcomes. Emerging techniques, including those driven by artificial intelligence, are rapidly expanding our ability to quantify the different domains of function during rehabilitation, other encounters with healthcare, and in the community. While this seems poised to usher rehabilitation into the era of big data and should be a powerful driver of precision rehabilitation, our field lacks a coherent framework to utilize these data and deliver on this promise. We propose a framework that builds upon multiple existing pillars to fill this gap. Our framework aims to identify the Optimal Dynamic Treatment Regimens (ODTR), or the decision-making strategy that takes in the range of available measurements and biomarkers to identify interventions likely to maximize long-term function. This is achieved by designing and fitting causal models, which extend the Computational Neurorehabilitation framework using tools from causal inference. These causal models can learn from heterogeneous data from different silos, which must include detailed documentation of interventions, such as using the Rehabilitation Treatment Specification System. The models then serve as digital twins of patient recovery trajectories, which can be used to learn the ODTR. Our causal modeling framework also emphasizes quantitatively linking changes across levels of the functioning to ensure that interventions can be precisely selected based on careful measurement of impairments while also being selected to maximize outcomes that are meaningful to patients and stakeholders. We believe this approach can provide a unifying framework to leverage growing big rehabilitation data and AI-powered measurements to produce precision rehabilitation treatments that can improve clinical outcomes.

In this paper, we consider the problem of causal order discovery within the framework of monotonic Structural Causal Models (SCMs), which have gained attention for their potential to enable causal inference and causal discovery from observational data. While existing approaches either assume prior knowledge about the causal order or use complex optimization techniques to impose sparsity in the Jacobian of Triangular Monotonic Increasing maps, our work introduces a novel sequential procedure that directly identifies the causal order by iteratively detecting the root variable. This method eliminates the need for sparsity assumptions and the associated optimization challenges, enabling the identification of a unique SCM without the need for multiple independence tests to break the Markov equivalence class. We demonstrate the effectiveness of our approach in sequentially finding the root variable, comparing it to methods that maximize Jacobian sparsity.

Many real-world systems can be usefully represented as sets of interacting components. Examples include computational systems, such as query processors and compilers, natural systems, such as cells and ecosystems, and social systems, such as families and organizations. However, current approaches to estimating potential outcomes and causal effects typically treat such systems as single units, represent them with a fixed set of variables, and assume a homogeneous data-generating process. In this work, we study a compositional approach for estimating individual-level potential outcomes and causal effects in structured systems, where each unit is represented by an instance-specific composition of multiple heterogeneous components. The compositional approach decomposes unit-level causal queries into more fine-grained queries, explicitly modeling how unit-level interventions affect component-level outcomes to generate a unit's outcome. We demonstrate this approach using modular neural network architectures and show that it provides benefits for causal effect estimation from observational data, such as accurate causal effect estimation for structured units, increased sample efficiency, improved overlap between treatment and control groups, and compositional generalization to units with unseen combinations of components. Remarkably, our results show that compositional modeling can improve the accuracy of causal estimation even when component-level outcomes are unobserved. We also create and use a set of real-world evaluation environments for the empirical evaluation of compositional approaches for causal effect estimation and demonstrate the role of composition structure, varying amounts of component-level data access, and component heterogeneity in the performance of compositional models as compared to the non-compositional approaches.

Energy markets exhibit complex causal relationships between weather patterns, generation technologies, and price formation, with regime changes occurring continuously rather than at discrete break points. Current approaches model electricity prices without explicit causal interpretation or counterfactual reasoning capabilities. We introduce Augmented Time Series Causal Models (ATSCM) for energy markets, extending counterfactual reasoning frameworks to multivariate temporal data with learned causal structure. Our approach models energy systems through interpretable factors (weather, generation mix, demand patterns), rich grid dynamics, and observable market variables. We integrate neural causal discovery to learn time-varying causal graphs without requiring ground truth DAGs. Applied to real-world electricity price data, ATSCM enables novel counterfactual queries such as "What would prices be under different renewable generation scenarios?".

Causal discovery is challenging in general dynamical systems because, without strong structural assumptions, the underlying causal graph may not be identifiable even from interventional data. However, many real-world systems exhibit directional, cascade-like structure, in which components activate sequentially and upstream failures suppress downstream effects. We study causal discovery in such chain-reaction systems and show that the causal structure is uniquely identifiable from blocking interventions that prevent individual components from activating. We propose a minimal estimator with finite-sample guarantees, achieving exponential error decay and logarithmic sample complexity. Experiments on synthetic models and diverse chain-reaction environments demonstrate reliable recovery from a few interventions, while observational heuristics fail in regimes with delayed or overlapping causal effects.

Causal Representation Learning (CRL) aims at identifying high-level causal factors and their relationships from high-dimensional observations, e.g., images. While most CRL works focus on learning causal representations in a single environment, in this work we instead propose a first step towards learning causal representations from temporal sequences of images that can be adapted in a new environment, or composed across multiple related environments. In particular, we introduce DECAF, a framework that detects which causal factors can be reused and which need to be adapted from previously learned causal representations. Our approach is based on the availability of intervention targets, that indicate which variables are perturbed at each time step. Experiments on three benchmark datasets show that integrating our framework with four state-of-the-art CRL approaches leads to accurate representations in a new environment with only a few samples.

Uncovering cause-effect relationships from observational time series is fundamental to understanding complex systems. While many methods infer static causal graphs, real-world systems often exhibit dynamic causality-where relationships evolve over time. Accurately capturing these temporal dynamics requires time-resolved causal graphs. We propose UnCLe, a novel deep learning method for scalable dynamic causal discovery. UnCLe employs a pair of Uncoupler and Recoupler networks to disentangle input time series into semantic representations and learns inter-variable dependencies via auto-regressive Dependency Matrices. It estimates dynamic causal influences by analyzing datapoint-wise prediction errors induced by temporal perturbations. Extensive experiments demonstrate that UnCLe not only outperforms state-of-the-art baselines on static causal discovery benchmarks but, more importantly, exhibits a unique capability to accurately capture and represent evolving temporal causality in both synthetic and real-world dynamic systems (e.g., human motion). UnCLe offers a promising approach for revealing the underlying, time-varying mechanisms of complex phenomena.

Judea Pearl's vision of Structural Causal Models (SCMs) as engines for counterfactual reasoning hinges on faithful abduction: the precise inference of latent exogenous noise. For decades, operationalizing this step for complex, non-linear mechanisms has remained a significant computational challenge. The advent of diffusion models, powerful universal function approximators, offers a promising solution. However, we argue that their standard design, optimized for perceptual generation over logical inference, introduces a fundamental flaw for this classical problem: an inherent information loss we term the Structural Reconstruction Error (SRE). To address this challenge, we formalize the principle of Causal Information Conservation (CIC) as the necessary condition for faithful abduction. We then introduce BELM-MDCM, the first diffusion-based framework engineered to be causally sound by eliminating SRE by construction through an analytically invertible mechanism. To operationalize this framework, a Targeted Modeling strategy provides structural regularization, while a Hybrid Training Objective instills a strong causal inductive bias. Rigorous experiments demonstrate that our Zero-SRE framework not only achieves state-of-the-art accuracy but, more importantly, enables the high-fidelity, individual-level counterfactuals required for deep causal inquiries. Our work provides a foundational blueprint that reconciles the power of modern generative models with the rigor of classical causal theory, establishing a new and more rigorous standard for this emerging field.

Causal discovery traditionally relies on statistical methods applied to observational data, often requiring large datasets and assumptions about underlying causal structures. Recent advancements in Large Language Models (LLMs) have introduced new possibilities for causal discovery by providing domain expert knowledge. However, it remains unclear whether LLMs can effectively process observational data for causal discovery. In this work, we explore the potential of LLMs for data-driven causal discovery by integrating observational data for LLM-based reasoning. Specifically, we examine whether LLMs can effectively utilize observational data through two prompting strategies: pairwise prompting and breadth first search (BFS)-based prompting. In both approaches, we incorporate the observational data directly into the prompt to assess LLMs' ability to infer causal relationships from such data. Experiments on benchmark datasets show that incorporating observational data enhances causal discovery, boosting F1 scores by up to 0.11 point using both pairwise and BFS LLM-based prompting, while outperforming traditional statistical causal discovery baseline by up to 0.52 points. Our findings highlight the potential and limitations of LLMs for data-driven causal discovery, demonstrating their ability to move beyond textual metadata and effectively interpret and utilize observational data for more informed causal reasoning. Our studies lays the groundwork for future advancements toward fully LLM-driven causal discovery.

Simulation-based testing is essential for evaluating the safety of Autonomous Driving Systems (ADSs). Comprehensive evaluation requires testing across diverse scenarios that can trigger various types of violations under different conditions. While existing methods typically focus on individual diversity metrics, such as input scenarios, ADS-generated motion commands, and system violations, they often fail to capture the complex interrelationships among these elements. This oversight leads to gaps in testing coverage, potentially missing critical issues in the ADS under evaluation. However, quantifying these interrelationships presents a significant challenge. In this paper, we propose a novel causality-aware fuzzing technique, Causal-Fuzzer, to enable efficient and comprehensive testing of ADSs by exploring causally diverse scenarios. The core of Causal-Fuzzer is constructing a causal graph to model the interrelationships among the diversities of input scenarios, ADS motion commands, and system violations. Then the causal graph will guide the process of critical scenario generation. Specifically, Causal-Fuzzer proposes (1) a causality-based feedback mechanism that quantifies the combined diversity of test scenarios by assessing whether they activate new causal relationships, and (2) a causality-driven mutation strategy that prioritizes mutations on input scenario elements with higher causal impact on ego action changes and violation occurrence, rather than treating all elements equally. We evaluated Causal-Fuzzer on an industry-grade ADS Apollo, with a high-fidelity. Our empirical results demonstrate that Causal-Fuzzer significantly outperforms existing methods in (1) identifying a greater diversity of violations, (2) providing enhanced testing sufficiency with improved coverage of causal relationships, and (3) achieving greater efficiency in detecting the first critical scenarios.

This study addresses the lack of structured causal modeling between tactical strike behavior and strategic delay in current strategic-level simulations, particularly the structural bottlenecks in capturing intermediate variables within the "resilience - nodal suppression - negotiation window" chain. We propose the Intervention-Aware Spatio-Temporal Graph Neural Network (IA-STGNN), a novel framework that closes the causal loop from tactical input to strategic delay output. The model integrates graph attention mechanisms, counterfactual simulation units, and spatial intervention node reconstruction to enable dynamic simulations of strike configurations and synchronization strategies. Training data are generated from a multi-physics simulation platform (GEANT4 + COMSOL) under NIST SP 800-160 standards, ensuring structural traceability and policy-level validation. Experimental results demonstrate that IA-STGNN significantly outperforms baseline models (ST-GNN, GCN-LSTM, XGBoost), achieving a 12.8 percent reduction in MAE and 18.4 percent increase in Top-5 percent accuracy, while improving causal path consistency and intervention stability. IA-STGNN enables interpretable prediction of strategic delay and supports applications such as nuclear deterrence simulation, diplomatic window assessment, and multi-strategy optimization, providing a structured and transparent AI decision-support mechanism for high-level policy modeling.

Despite their strong performance on reasoning benchmarks, large language models (LLMs) have proven brittle when presented with counterfactual questions, suggesting weaknesses in their causal reasoning ability. While recent work has demonstrated that labeled counterfactual tasks can be useful benchmarks of LLMs' causal reasoning, producing such data at the scale required to cover the vast potential space of counterfactuals is limited. In this work, we introduce double counterfactual consistency (DCC), a lightweight inference-time method for measuring and guiding the ability of LLMs to reason causally. Without requiring labeled counterfactual data, DCC verifies a model's ability to execute two important elements of causal reasoning: causal intervention and counterfactual prediction. Using DCC, we evaluate the causal reasoning abilities of various leading LLMs across a range of reasoning tasks and interventions. Moreover, we demonstrate the effectiveness of DCC as a training-free test-time rejection sampling criterion and show that it can directly improve performance on reasoning tasks across multiple model families.

Counterfactual reasoning has emerged as a crucial technique for generalizing the reasoning capabilities of large language models (LLMs). By generating and analyzing counterfactual scenarios, researchers can assess the adaptability and reliability of model decision-making. Although prior work has shown that LLMs often struggle with counterfactual reasoning, it remains unclear which factors most significantly impede their performance across different tasks and modalities. In this paper, we propose a decompositional strategy that breaks down the counterfactual generation from causality construction to the reasoning over counterfactual interventions. To support decompositional analysis, we investigate \ntask datasets spanning diverse tasks, including natural language understanding, mathematics, programming, and vision-language tasks. Through extensive evaluations, we characterize LLM behavior across each decompositional stage and identify how modality type and intermediate reasoning influence performance. By establishing a structured framework for analyzing counterfactual reasoning, this work contributes to the development of more reliable LLM-based reasoning systems and informs future elicitation strategies.

Large language models (LLMs) are increasingly used in domains where causal reasoning matters, yet it remains unclear whether their judgments reflect normative causal computation, human-like shortcuts, or brittle pattern matching. We benchmark 20+ LLMs against a matched human baseline on 11 causal judgment tasks formalized by a collider structure ($C_1 \rightarrow E \leftarrow C_2$). We find that a small interpretable model compresses LLMs' causal judgments well and that most LLMs exhibit more rule-like reasoning strategies than humans who seem to account for unmentioned latent factors in their probability judgments. Furthermore, most LLMs do not mirror the characteristic human collider biases of weak explaining away and Markov violations. We probe LLMs' causal judgment robustness under (i) semantic abstraction and (ii) prompt overloading (injecting irrelevant text), and find that chain-of-thought (CoT) increases robustness for many LLMs. Together, this divergence suggests LLMs can complement humans when known biases are undesirable, but their rule-like reasoning may break down when uncertainty is intrinsic - highlighting the need to characterize LLM reasoning strategies for safe, effective deployment.

Causality is vital for understanding true cause-and-effect relationships between variables within predictive models, rather than relying on mere correlations, making it highly relevant in the field of Explainable AI. In an automated decision-making scenario, causal inference methods can analyze the underlying data-generation process, enabling explanations of a model's decision by manipulating features and creating counterfactual examples. These counterfactuals explore hypothetical scenarios where a minimal number of factors are altered, providing end-users with valuable information on how to change their situation. However, interpreting a set of multiple counterfactuals can be challenging for end-users who are not used to analyzing raw data records. In our work, we propose a novel multi-step pipeline that uses counterfactuals to generate natural language explanations of actions that will lead to a change in outcome in classifiers of tabular data using LLMs. This pipeline is designed to guide the LLM through smaller tasks that mimic human reasoning when explaining a decision based on counterfactual cases. We conducted various experiments using a public dataset and proposed a method of closed-loop evaluation to assess the coherence of the final explanation with the counterfactuals, as well as the quality of the content. Results are promising, although further experiments with other datasets and human evaluations should be carried out.

Structural Equation Models (SEM) are the standard approach to representing causal dependencies between variables in causal models. In this paper we propose a new interpretation of SEMs when reasoning about Actual Causality, in which SEMs are viewed as mechanisms transforming the dynamics of exogenous variables into the dynamics of endogenous variables. This allows us to combine counterfactual causal reasoning with existing temporal logic formalisms, and to introduce a temporal logic, CPLTL, for causal reasoning about such structures. We show that the standard restriction to so-called \textit{recursive} models (with no cycles in the dependency graph) is not necessary in our approach, allowing us to reason about mutually dependent processes and feedback loops. Finally, we introduce new notions of model equivalence for temporal causal models, and show that CPLTL has an efficient model-checking procedure.

Hierarchical reinforcement learning (HRL) improves the efficiency of long-horizon reinforcement-learning tasks with sparse rewards by decomposing the task into a hierarchy of subgoals. The main challenge of HRL is efficient discovery of the hierarchical structure among subgoals and utilizing this structure to achieve the final goal. We address this challenge by modeling the subgoal structure as a causal graph and propose a causal discovery algorithm to learn it. Additionally, rather than intervening on the subgoals at random during exploration, we harness the discovered causal model to prioritize subgoal interventions based on their importance in attaining the final goal. These targeted interventions result in a significantly more efficient policy in terms of the training cost. Unlike previous work on causal HRL, which lacked theoretical analysis, we provide a formal analysis of the problem. Specifically, for tree structures and, for a variant of Erdős-Rényi random graphs, our approach results in remarkable improvements. Our experimental results on HRL tasks also illustrate that our proposed framework outperforms existing work in terms of training cost.

Feature selection aims to preprocess the target dataset, find an optimal and most streamlined feature subset, and enhance the downstream machine learning task. Among filter, wrapper, and embedded-based approaches, the reinforcement learning (RL)-based subspace exploration strategy provides a novel objective optimization-directed perspective and promising performance. Nevertheless, even with improved performance, current reinforcement learning approaches face challenges similar to conventional methods when dealing with complex datasets. These challenges stem from the inefficient paradigm of using one agent per feature and the inherent complexities present in the datasets. This observation motivates us to investigate and address the above issue and propose a novel approach, namely HRLFS. Our methodology initially employs a Large Language Model (LLM)-based hybrid state extractor to capture each feature's mathematical and semantic characteristics. Based on this information, features are clustered, facilitating the construction of hierarchical agents for each cluster and sub-cluster. Extensive experiments demonstrate the efficiency, scalability, and robustness of our approach. Compared to contemporary or the one-feature-one-agent RL-based approaches, HRLFS improves the downstream ML performance with iterative feature subspace exploration while accelerating total run time by reducing the number of agents involved.

Safe reinforcement learning (RL) is crucial for real-world applications, and multi-agent interactions introduce additional safety challenges. While Probabilistic Logic Shields (PLS) has been a powerful proposal to enforce safety in single-agent RL, their generalizability to multi-agent settings remains unexplored. In this paper, we address this gap by conducting extensive analyses of PLS within decentralized, multi-agent environments, and in doing so, propose $\textbf{Shielded Multi-Agent Reinforcement Learning (SMARL)}$ as a general framework for steering MARL towards norm-compliant outcomes. Our key contributions are: (1) a novel Probabilistic Logic Temporal Difference (PLTD) update for shielded, independent Q-learning, which incorporates probabilistic constraints directly into the value update process; (2) a probabilistic logic policy gradient method for shielded PPO with formal safety guarantees for MARL; and (3) comprehensive evaluation across symmetric and asymmetrically shielded $n$-player game-theoretic benchmarks, demonstrating fewer constraint violations and significantly better cooperation under normative constraints. These results position SMARL as an effective mechanism for equilibrium selection, paving the way toward safer, socially aligned multi-agent systems.

To build intelligent machine learning systems, there are two broad approaches. One approach is to build inherently interpretable models, as endeavored by the growing field of causal representation learning. The other approach is to build highly-performant foundation models and then invest efforts into understanding how they work. In this work, we relate these two approaches and study how to learn human-interpretable concepts from data. Weaving together ideas from both fields, we formally define a notion of concepts and show that they can be provably recovered from diverse data. Experiments on synthetic data and large language models show the utility of our unified approach.

We present a generative optimization approach for learning game-playing agents, where policies are represented as Python programs and refined using large language models (LLMs). Our method treats decision-making policies as self-evolving code, with current observation as input and an in-game action as output, enabling agents to self-improve through execution traces and natural language feedback with minimal human intervention. Applied to Atari games, our game-playing Python program achieves performance competitive with deep reinforcement learning (RL) baselines while using significantly less training time and much fewer environment interactions. This work highlights the promise of programmatic policy representations for building efficient, adaptable agents capable of complex, long-horizon reasoning.

Causal games are probabilistic graphical models that enable causal queries to be answered in multi-agent settings. They extend causal Bayesian networks by specifying decision and utility variables to represent the agents' degrees of freedom and objectives. In multi-agent settings, whether each agent decides on their policy before or after knowing the causal intervention is important as this affects whether they can respond to the intervention by adapting their policy. Consequently, previous work in causal games imposed chronological constraints on permissible interventions. We relax this by outlining a sound and complete set of primitive causal interventions so the effect of any arbitrarily complex interventional query can be studied in multi-agent settings. We also demonstrate applications to the design of safe AI systems by considering causal mechanism design and commitment.

We study a specific type of SCM, called a Dynamic Structural Causal Model (DSCM), whose endogenous variables represent functions of time, which is possibly cyclic and allows for latent confounding. As a motivating use-case, we show that certain systems of Stochastic Differential Equations (SDEs) can be appropriately represented with DSCMs. An immediate consequence of this construction is a graphical Markov property for systems of SDEs. We define a time-splitting operation, allowing us to analyse the concept of local independence (a notion of continuous-time Granger (non-)causality). We also define a subsampling operation, which returns a discrete-time DSCM, and which can be used for mathematical analysis of subsampled time-series. We give suggestions how DSCMs can be used for identification of the causal effect of time-dependent interventions, and how existing constraint-based causal discovery algorithms can be applied to time-series data.

Large language models (LLMs) have achieved success in acting as agents, which interact with environments through tools such as search engines. However, LLMs are optimized for language generation instead of tool use during training or alignment, limiting their effectiveness as agents. To resolve this problem, previous work has first collected interaction trajectories between LLMs and environments, using only trajectories that successfully finished the task to fine-tune smaller models, making fine-tuning data scarce and acquiring it both difficult and costly. Discarding failed trajectories also leads to significant wastage of data and resources and limits the possible optimization paths during fine-tuning. In this paper, we argue that unsuccessful trajectories offer valuable insights, and LLMs can learn from these trajectories through appropriate quality control and fine-tuning strategies. By simply adding a prefix or suffix that tells the model whether to generate a successful trajectory during training, we improve model performance by a large margin on mathematical reasoning, multi-hop question answering, and strategic question answering tasks. We further analyze the inference results and find that our method provides a better trade-off between valuable information and errors in unsuccessful trajectories. To our knowledge, we are the first to demonstrate the value of negative trajectories and their application in agent-tunning scenarios. Our findings offer guidance for developing better agent-tuning methods and low-resource data usage techniques.

Agent-Based Model (ABM) validation is crucial as it helps ensuring the reliability of simulations, and causal discovery has become a powerful tool in this context. However, current causal discovery methods often face accuracy and robustness challenges when applied to complex and noisy time series data, which is typical in ABM scenarios. This study addresses these issues by proposing a Robust Cross-Validation (RCV) approach to enhance causal structure learning for ABM validation. We develop RCV-VarLiNGAM and RCV-PCMCI, novel extensions of two prominent causal discovery algorithms. These aim to reduce the impact of noise better and give more reliable causal relation results, even with high-dimensional, time-dependent data. The proposed approach is then integrated into an enhanced ABM validation framework, which is designed to handle diverse data and model structures. The approach is evaluated using synthetic datasets and a complex simulated fMRI dataset. The results demonstrate greater reliability in causal structure identification. The study examines how various characteristics of datasets affect the performance of established causal discovery methods. These characteristics include linearity, noise distribution, stationarity, and causal structure density. This analysis is then extended to the RCV method to see how it compares in these different situations. This examination helps confirm whether the results are consistent with existing literature and also reveals the strengths and weaknesses of the novel approaches. By tackling key methodological challenges, the study aims to enhance ABM validation with a more resilient valuation framework presented. These improvements increase the reliability of model-driven decision making processes in complex systems analysis.

Synthetic tabular data generation addresses data scarcity and privacy constraints in a variety of domains. Tabular Prior-Data Fitted Network (TabPFN), a recent foundation model for tabular data, has been shown capable of generating high-quality synthetic tabular data. However, TabPFN is autoregressive: features are generated sequentially by conditioning on the previous ones, depending on the order in which they appear in the input data. We demonstrate that when the feature order conflicts with causal structure, the model produces spurious correlations that impair its ability to generate synthetic data and preserve causal effects. We address this limitation by integrating causal structure into TabPFN's generation process through two complementary approaches: Directed Acyclic Graph (DAG)-aware conditioning, which samples each variable given its causal parents, and a Completed Partially Directed Acyclic Graph (CPDAG)-based strategy for scenarios with partial causal knowledge. We evaluate these approaches on controlled benchmarks and six CSuite datasets, assessing structural fidelity, distributional alignment, privacy preservation, and Average Treatment Effect (ATE) preservation. Across most settings, DAG-aware conditioning improves the quality and stability of synthetic data relative to vanilla TabPFN. The CPDAG-based strategy shows moderate improvements, with effectiveness depending on the number of oriented edges. These results indicate that injecting causal structure into autoregressive generation enhances the reliability of synthetic tabular data.