Physical Reasoning on InterPhyre Down to Earth (evaluation)
72AccuracyHEXA
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| HEXABackbone=Qwen-2.5-7B, Search Configuration=Online-Evolving (k=3 seeds per round, 17 rounds), Seeds=502026.06 | 72 | 12.6 | 10 | |
| REACT baselineBackbone=Qwen-2.5-7B, Search Configuration=Online-Evolving (k=3 seeds per round, 17 rounds), Seeds=502026.06 | 62 | 12.5 | — | |
| HEXABackbone=Qwen-2.5-3B, Search Configuration=Online-Evolving (k=3 seeds per round, 17 rounds), Seeds=502026.06 | 24 | 18.6 | 16 | |
| REACT baselineBackbone=Qwen-2.5-3B, Search Configuration=Online-Evolving (k=3 seeds per round, 17 rounds), Seeds=502026.06 | 8 | 23.8 | — |