Code Review for Optimization Papers
We review optimization papers and their accompanying implementations for paper-to-code consistency, test whether experiments support the stated claims, and assess benchmark fairness and reproducibility.
Findings across 50 papers
Optimend found 1,105 issues in a sample from Mathematical Programming Computation and Mathematics of Operations Research. Each finding is assigned to one category.
- Mathematical claims 264
- Algorithms and code 206
- Agreement between paper and code 162
- Files needed to reproduce results 107
- Numerical error and tolerances 64
- Reported results and metrics 62
- Time, memory, and compute 57
- Experiment design and uncertainty 54
- Baselines and test coverage 54
- Citations and prior work 38
- Reporting and claim scope 37
Review Cases
Citation
If you find the feedback from this webapp useful in your paper, please cite:
OptiMend: Auditing Optimization Papers with Agentic AI
Wanyu Zhang*, Angikar Ghosal*, Madeleine Udell*= co-first authors.