| Public Physics Benchmarks (GPQA, SciBench, PhysReason) (test) | Reason, Reward, Refine | Average Accuracy64.4 | | 41 | 18d ago |
| MMLU-Pro Physics | SC | Accuracy88.5 | | 20 | 16d ago |
| MemoryArena Physics | OPD-Evolver-9B | Success Metric11.63 | | 20 | 1mo ago |
| Physics C-Eval WebInstruct | Qwen3-4B-Base (HEAL) | C-Eval Score80.15 | | 18 | 3mo ago |
| OlyBench Phy | TTS | Acceptance Length4.5 | | 12 | 2mo ago |
| IPhO Mechanics | Qwen3-30B + RL (synthetic) | Accuracy40 | | 10 | 3mo ago |
| HCV | Qwen3-30B + RL (synthetic) | Accuracy59 | | 10 | 3mo ago |
| Synthetic Symbolic | Qwen2.5-32B + RL (synthetic) | Accuracy10.4 | | 10 | 3mo ago |
| Synthetic Numeric | Qwen2.5-32B + RL (synthetic) | Accuracy21.9 | | 10 | 3mo ago |
| QFT Synthetic Hard | | Accuracy44.5 | | 6 | 3mo ago |
| QFT Synthetic Medium | | Accuracy88.8 | | 6 | 3mo ago |
| QFT Synthetic Easy | | Accuracy89 | | 6 | 3mo ago |
| arXiv human-adapted | | Accuracy (hep-th)60.4 | | 6 | 3mo ago |
| Human-Adapted (Pedagogical) | | Accuracy (Textbooks)61.4 | | 6 | 3mo ago |
| PhyBench | Ministral 3 | pass@1626 | | 6 | 4mo ago |
| PhysReason | ARYA | Overall Score73.3 | | 4 | 4mo ago |
| Physics Puzzles Two Body Problem (test) | HExA | Accuracy34 | | 2 | 25d ago |
| Physics Puzzles Falling Into Place (test) | HExA | Accuracy32 | | 2 | 25d ago |
| Physics Puzzles Catapult (test) | HExA | Accuracy44 | | 2 | 25d ago |
| PHYSICS | Qwen2.5 32B + RL (synthetic) | Mean Accuracy43.09 | | 2 | 3mo ago |