How to Write a Mathematical Modeling Paper: Format, Judging Criteria, and Templates
Acadwrite Team · Last updated July 24, 2026
Official 2025 award data: 68,311 teams from 1,837 institutions competed, and 294 undergraduate teams won a national first prize (0.48%). Accounts from judges and winners suggest that an initial review takes roughly 5–10 minutes per paper. This guide explains how judges read, what each section must contain, and which format errors can disqualify a submission.
How Judges Read Your Paper
The national contest uses two review levels. Regional committees review first, and regional awards generally go to no more than one third of teams. Strong regional submissions then enter national review (see the official national review procedures). At least two groups of judges therefore read your paper, and the first reads quickly.
What do they judge? The competition charter reduces it to one sentence:
“Reasonable assumptions, creative modeling, correct results, and clear writing.”
Each phrase maps to concrete checks inside the paper:
| Official criterion | Where it appears in the paper |
|---|---|
| Reasonable assumptions | Every assumption should identify the model step it simplifies. If it has no corresponding role, a judge will treat it as filler. |
| Creative modeling | Show that the model matches the problem: why this decision variable was chosen and which problem condition each constraint represents. |
| Correct results | Provide verifiable numbers with units and plausible scale, then use sensitivity analysis to show that the conclusion was not accidental. |
| Clear writing | A fast-reading judge should be able to follow abstract → figures → conclusion and retell what you did. If not, the writing is unclear. |
Initial Review Is Brief, So the Abstract Must Present the Full Result
The following three claims come from judges' and winners' experience rather than official rules. Together they show the abstract's importance in initial review.
「The first round reads only the abstract」
「The abstract and results largely decide whether a paper receives provincial first prize or second prize」
「The abstract separates participation awards from provincial awards」
Initial review takes roughly 5–10 minutes per paper. The abstract must state the problem, model, main numerical results, and conclusion directly. The next section shows how.
Four Abstract Elements: State the Model, Numbers, and Conclusion
The official format gives the abstract one full page, including the title and keywords. Answer four questions in order:
- 1What problem did you solve? — Which subproblem, what input data, and what requested output?
- 2Which model and method did you use? — Name the model type and solution algorithm without deriving them here.
- 3Which numerical results did you obtain? — Give the key number and unit for every subproblem.
- 4What do the results mean? — State the conclusion behind the numbers and whether sensitivity analysis shows it is stable.
The difference is clearest in an example (invented solely for illustration):
“For Problem 1, we developed a reasonable mathematical model and used an effective algorithm to obtain satisfactory results. For Problem 2, we improved the previous model; the results have practical significance and provide a useful reference.”
This answers only half of the four elements: it names a method without defining the problem, gives no numbers, and states no conclusion.
“For Problem 1, we used scheduling order as the decision variable in an integer program. Branch and bound produced an optimum total cost of RMB 18.42 million, 12.6% below the greedy baseline. Sensitivity analysis showed that the optimal structure remained unchanged when unit freight cost rose 10%, indicating a robust conclusion.”
Problem, model, algorithm, number, conclusion, and robustness are all present and verifiable.
Section-by-Section Checks and Common Problems
Each body section has a distinct task. Check them one by one using the standards below.
Problem restatement
Reorganize the task in your own words: supplied data, constraints, and required output for every subproblem.
Copying the prompt. The judge just read it, so repetition signals that this section can be skipped.
Model assumptions
Connect every assumption to a specific model simplification and show where it takes effect in the modeling section.
Generic assumptions such as “the data are reliable” or “other factors are ignored” that have no role in the model.
Modeling and solution
Number and title every figure and table and cite it explicitly in the prose, such as “Figure 3 shows ...”. Define every symbol at first use.
Unreferenced figures leave judges unsure where to look, while uninterrupted equations provide no narrative connection.
Sensitivity analysis
Perturb at least one key parameter, report the resulting numbers, and state whether the conclusion changes.
Writing only “testing shows that the model is stable” without any numerical evidence.
References
Use complete scientific citation records and ensure every in-text citation corresponds to a bibliography entry.
Padding the list with inconsistently formatted links from unclear sources.
For exact bibliography formats, use our reference-format guideas a line-by-line checklist.
Format Checklist
The following items are distilled from the official Paper Format Specification (2025 revision). A 2026 revisionhas now been published; always check the original revision for your competition year before submission.
- ✓A4 paper, margins of at least 2.5 cm on all sides, bound on the left
- ✓Page 1 is the declaration and page 2 the identification-number page; use the official format without changes
- ✓Page 3 is the abstract page with title and keywords, limited to one page; page numbering starts at 1
- ✓The body starts on page 4; omit a table of contents and keep the body within 20 pages where possible
- Appendices have no page limit but must contain all complete, runnable source code; missing or non-runnable code disqualifies the paper
- The full paper is anonymous: neither body nor appendices may contain a university, name, or other identity information
- ✓Submit one PDF no larger than 20 MB, with the abstract as its first page
- ✓Package supporting files as a ZIP no larger than 20 MB
Appendix source code and anonymity directly affect award eligibility. Run all complete appendix code before submission and search the paper and supporting files for university names, personal names, and other identity information.
Choosing a LaTeX Template: Recent Maintenance Before Stars
Some online templates still use the 2023 format. Their declaration, identification page, or AI-tool report may not match the current rules. We checked the maintenance status of popular GitHub templates in July 2026:
| Repository | Stars | Status | Notes |
|---|---|---|---|
| EmpyreanHYR/CUMCM-Latex-template | — | 2026-06 | Actively maintained and includes the AI-tool usage report required by the 2025 rules |
| Sustainable-Enjoyment/CUMCM-LaTeX-Template | — | 2025-08 | Maintained by a national first-prize winner |
| latexstudio/CUMCMThesis | 940★ | Format updated through 2023 | A widely circulated legacy template that must be revised against the current rules |
| personqianduixue/Math_Model | 4617★ | Resource collection | A collection of paper templates, modeling code, and study resources |
The qyxf/mcm-cumcm-thesis repository still appears in search results despite being unmaintained for seven years and should not be used. Whatever template you choose, verify it against the current revision using the checklist above. The team remains responsible for compliance.
Award-Winning Papers and Problem Reviews
Award-winning papers for the same problem show a complete implementation. Problem-author reviews explain design intent and judging priorities. These three official entry points provide free access:
We aggregate entry points only; the original papers remain on official sites. These links were checked in July 2026.
Turn Problem Sources and Computed Results into a Paper
Acadwrite can search and read background and method sources for the problem, analyze data and model results, and create required tables and figures. Once the modeling approach is confirmed, it drafts the paper around restatement, assumptions, model, solution, validation, and references.
Frequently Asked Questions
How long should the abstract be?
The official format limits the dedicated abstract page, including title and keywords, to one page. Fill it with the four elements—problem, method, numerical result, and conclusion. Every subproblem's key number should appear instead of a vague claim that results were satisfactory.
Can we submit if we did not finish every subproblem?
Yes. State honestly which problems were completed and which numerical results were obtained. Present only completed, verified content, and ensure appendix code matches the paper and runs successfully.
Should a mathematical modeling paper have a table of contents?
No. The official format starts the body on page 4 and recommends keeping it within 20 pages. Remove the table of contents when using an old template.
Can teams use AI tools?
The 2025 revision requires an AI-tool usage report and honest disclosure. Some maintained templates include the report format. Teams must still solve the problem independently and may not discuss it with outsiders or join online problem-discussion groups.
Continue Reading
Turn Your Modeling Approach into a Complete Paper with Acadwrite
Search and read background sources, analyze data and model results, create figures, and draft the paper in the competition's required structure.