AI paper review
Get past the reviewer before you submit: novelty claims checked against the literature, comments written to Nature reviewing standards, and revisions for the ones you pick.
Paste text or upload .docx / .txt / .md. Credits are charged by actual usage. See credit details
How AI paper review works
- 1
Submit your paper
Paste the text or upload a file; add target journal, course or conference requirements if you have them.
- 2
Check novelty claims
Claims of "first" or "novel" are searched against prior work one by one; nothing is called prior art without evidence.
- 3
Write the review
Reviewed to Nature standards; every comment quotes the text, names its category and gives an actionable fix.
- 4
Pick and revise
Choose the comments to address and the text is revised for those items only.
Finds problems like a real reviewer, then helps fix them
Comments come with quotes and evidence, and revisions land only where you choose.
Novelty claims checked against the literature
Each claimed contribution is searched against prior work: has someone published the same or a very similar method, or is this a small change to an existing framework? Overlaps come with the papers found.
Reviewed to Nature standards
It assesses originality, scientific importance, interdisciplinary readership, technical soundness and readability for nonspecialists, separates major and minor concerns, and never pads the list.
Every comment points to the text
Each comment quotes the passage it is about and says what the problem is, why it matters and how to fix it, so you can find it at once instead of reading vague advice.
Revised only where you choose
The review is a checklist. Pick the items to address and the text changes for those only, with nothing else touched and existing citations kept.
Watch a pre-submission review
From submitting an abstract to revising for chosen comments, step by step in the workspace. The abstract is written to show the process; the cited paper is real.
We propose, for the first time, a dynamic token pruning methodBuilding on DynamicViT (Rao et al., 2021), we propose a dynamic token pruning method that prunes tokens at adaptive ratios, achieving a revolutionary speed-up on ImageNet with 37% faster inferencereaching 37% faster inference on ImageNet, and far outperforming existing methods on all metricsoutperforming the baselines at equal Top-1 accuracy.
Struck-through text is removed; highlighted text is the revision.
Review this before I submit to CVPR
Checking novelty claims
dynamic token pruning vision transformerDynamicViT (Rao et al., 2021, NeurIPS)adaptive token sparsification ViT3 related works found
The core contribution overlaps with prior work, key conclusions lack statistical support, and method details are insufficient to reproduce.
Review comments4 items
- OriginalityDynamic token pruning is not new
We propose, for the first time, a dynamic token pruning method
DynamicViT already prunes redundant tokens dynamically by input (Rao et al., 2021). Cite it and explain the difference; until then, drop "for the first time".
- Argument"Far outperforming" lacks statistics
far outperforming existing methods on all metrics
Only single-run means are given. Report standard deviations or significance tests across runs, and soften the claim if results are not significant.
- ReproducibilityPruning ratios and training setup missing
prunes tokens at adaptive ratios
State the range of keep ratios per layer, training epochs and learning rate; otherwise it cannot be reproduced.
- Wording"Revolutionary" is unsupported
achieving a revolutionary speed-up
Replace with what the data supports, such as the measured inference speed-up.
Revised items 1, 2, 4.
After the review, keep going
Reviewed against your journal or course
Add your target journal's author guidelines or a course or conference style guide as extra requirements, and they are assessed too.
Missing citations come with sources
When challenging novelty or flagging missing or wrong citations, comments cite real papers found in the search; other comments are not padded with references.
Polish once revised
After revising for the review, keep polishing wording and adding citations in the same workspace.
Paper polishingAcknowledged in published SCI papers
Published papers state in their acknowledgments or AI-use statements that Acadwrite was used to polish the language.
- Theranostics2024Microfluidic chips in female reproduction: a systematic review of status, advances, and challenges
During the preparation of this work the authors used Acadwrite in order to improve readability, grammar and language style.
- Frontiers in Neuroscience2024Physical exercise for brain plasticity promotion an overview of the underlying oscillatory mechanism
We would like to thank Acadwrite (acadwrite.cn) for English language editing.
- Poultry Science2025Effects of induced molting on lipid accumulation in liver of aged laying hens
During the preparation of this work authors used (https://acadwrite.cn/) in order to improve language.
- Aging Medicine2025The Mechanism and Potential Therapeutic Strategies of Vascular Aging
Acadwrite AI (https://acadwrite.cn/apps/polish) was utilized for polishing the English language of the manuscript to enhance readability.
- Interactive Journal of Medical Research2025Evolution of Artificial Intelligence in Medical Education From 2000 to 2024: Bibliometric Analysis
During the preparation of this work, the authors used Acadwrite to improve readability, grammar, and language style.
AI paper review FAQ
Didn't find your answer? Reviews are charged by actual usage; see credit details. See credit details
What standard does the AI review use?
Nature's reviewing criteria: originality, scientific importance, interdisciplinary readership, technical soundness and readability, with major and minor concerns separated.
Will the AI just say nice things?
No flattery, and no padding the list with invented problems. Severity reflects impact on the paper's central case, and the wording is direct but professional.
How are novelty claims checked?
Claims of "first" or "novel" are turned into searches of prior work and compared against full abstracts. When the same or a very similar method exists, the comment cites the papers found; nothing is called prior art without evidence.
Can it revise the paper for me?
Yes. The review is a checklist; pick the items and the text is revised for those only. Items that need new experiments or data are skipped with an explanation.
Can it review against my target journal?
Yes. Add the journal's author guidelines or a course or conference style guide as extra requirements.
Does it support English papers?
Yes. Chinese and English papers are both supported, and comments follow your paper's language.
How much does it cost?
Sign-in is required. Reviews run in the workspace and are charged by tokens used, at 20 credits per million tokens. See the credits page for details.