Physics Problem Solving on PHYSICS (held-out)
20.8DBSEXPERT
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| EXPERTModel=Qwen3-1.7B optimized2026.02 | 20.8 | — | |
| DataChef-32BOracle Upper Bound=true, Samples=322026.02 | 10.4 | — | |
| Gemini-3-Pro2026.02 | 9.2 | 69.7 | |
| Kimi-K22026.02 | 9 | 19.7 | |
| Qwen3-Next ⊕ Kimi-K2Reasoning backbone=Qwen3-Next-80B, Coding backbone=Kimi-K2-Instruct2026.02 | 8.9 | 48.7 | |
| DataChef-32B2026.02 | 8.7 | 61.4 | |
| SOURCEbest2026.02 | 8.5 | — | |
| SOURCEavg2026.02 | 6.1 | — | |
| Qwen3-32B2026.02 | 5.9 | 11 |