Competition Math Grading on GraderBench 1000 datapoints
54Accuracyo3
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| o3Prompt type=ProofAutoGrader, Reference solution availability=No2026.02 | 54 | 20.2 | |
| Gemini 2.5 Deep ThinkPrompt type=ProofAutoGrader, Reference solution availability=No2026.02 | 52.5 | 20.5 | |
| Gemini 2.5 Deep ThinkPrompt type=ProofAutoGrader, Reference solution availability=No, Model variant=IMO Gold2026.02 | 50.2 | 18.4 | |
| MomusPrompt type=Momus, Reference solution availability=No, Pipeline solutions usage=Yes2026.02 | 49.8 | 19.9 | |
| o4-mini (high reasoning)Prompt type=ProofAutoGrader, Reference solution availability=No2026.02 | 47.3 | 25.2 | |
| Gemini 2.5 ProPrompt type=ProofAutoGrader, Reference solution availability=No2026.02 | 44.3 | 30.2 | |
| MomusPrompt type=Momus, Reference solution availability=No2026.02 | 28.1 | 25.9 |