Scientific Reasoning on PHYBench
30.34AccuracyInftyThink+
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| InftyThink+RL Setting=Task reward only (✓ T), Sampling Temperature=0.7, Number of Samples=322026.02 | 30.34 | 30.34 | 219.01 | |
| InftyThink+RL Setting=Task and efficiency reward (✓ T+E), Sampling Temperature=0.7, Number of Samples=322026.02 | 29.62 | 14.63 | 97.57 | |
| VanillaRL Setting=Task reward only (✓ T), Sampling Temperature=0.7, Number of Samples=322026.02 | 20.11 | 20.46 | 320.06 | |
| InftyThink+RL Setting=Cold start only (✗), Sampling Temperature=0.7, Number of Samples=322026.02 | 19.14 | 14.6 | 94.55 | |
| VanillaRL Setting=Cold start only (✗), Sampling Temperature=0.7, Number of Samples=322026.02 | 16.25 | 13.44 | 149.38 |