latex-paper-review

LaTeX Paper Review

Overview

Produce one concrete, compileable LaTeX review file. Prioritize scientific and technical correctness over prose polish, then add bounded editorial findings that materially improve clarity or professionalism.

Default Behavior

1. Read the full manuscript before writing findings.
2. Use extra thinking by default, but do not narrate the reasoning process.
3. Write a standalone LaTeX review to paper-reviews/review-YYYY-MM-DD-HHMMSS.tex, unless the user gives a different path.
4. Match the user's requested output language; default to Simplified Chinese (简体中文) unless specified otherwise.
5. Preserve equations, symbols, citation keys, and quoted manuscript snippets exactly when needed for technical accuracy.
6. Do not require a fixed review template or section layout unless the user asks for one.
7. Create a new timestamped review file by default. Only update an existing review file if the user explicitly asks.
8. Compiling to PDF is optional unless the user asks for it or compilation is needed to validate the LaTeX.

Review Priority

Spend attention in this order:

1. equations, derivations, notation, units, definitions, algorithms, and implementation-facing formulas
2. consistency between claims, evidence, figures, tables, captions, appendices, and conclusions
3. editorial quality, especially high-confidence issues that are cheap to verify and fix

Do not let editorial findings crowd out technical findings. Treat proofread, flag typos, grammar, writing quality, and similar wording as a request to spend more budget on the editorial sweep after the technical audit.

Workflow

1. Read the manuscript to understand the main claims, evidence, equations, appendices, and structure.
2. Build a structured internal map of the manuscript:
   • Research Question Map: Identify core research questions, primary claims, and intended contributions.
   • Construct / Variable / Parameter Map: Identify independent/dependent variables, mathematical parameters, system components, or theoretical constructs (relevant to the paper type).
   • Relationship Map: Identify hypothesized directions, logical dependencies, mathematical relationships, or system interactions.
   • Method Fit Assessment: Evaluate whether the selected research design or proof strategy logically answers the research questions, and whether conclusion boundaries match the evidence.
3. Trace the Evidence Chain for identified claims: Check the path Claim -> Evidence -> Figure/Table -> Method -> Citation -> Appendix (if applicable) for inconsistencies, missing links, or overreaching assertions.
4. Identify the highest-risk technical content: key equations, symbol definitions, quantitative claims, coordinate transformations, appendix formulas, and implementation-facing expressions.
5. Perform the technical audit, checking specific methodological validity dimensions (such as selection bias, endogeneity, validity, and boundary conditions where applicable).
6. Re-check the findings and classify them by severity and certainty before presenting.
7. Perform a bounded editorial pass after the technical audit.
8. Run scripts/proofing_scan.py when code execution is available; otherwise do the manual pattern scan in Final Proofing Sweep.
9. Spot-check proofing-scan candidates before presenting them as issues.
10. Write the review file.