Reasoning on MATH-500 (Error, ErrorGap, STP %)
2.57Error RateG-PAC
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| G-PACScoring method=Verbalized score, Alpha=0.05, Epsilon=0.052026.01 | 2.57 | 0 | 41.63 | |
| G-PACScoring method=Router-based score, Alpha=0.05, Epsilon=0.052026.01 | 2.63 | 0 | 49.3 | |
| G-PACScoring method=Logits-based score, Alpha=0.05, Epsilon=0.052026.01 | 2.78 | 0 | 43.44 | |
| PACScoring method=Logits-based score, Alpha=0.05, Epsilon=0.052026.01 | 3.08 | 1.02 | 56.86 | |
| PACScoring method=Verbalized score, Alpha=0.05, Epsilon=0.052026.01 | 3.24 | 0.11 | 61.14 | |
| PACScoring method=Router-based score, Alpha=0.05, Epsilon=0.052026.01 | 3.26 | 1.63 | 59.2 |