Mathematical Reasoning on ProcessBench (PB)
0.766AUCCPMI_Merge
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| CPMI_MergeReward Type=CPMI_Merge2026.04 | 0.766 | 36.8 | 3.38 | 4.45 | |
| CPMIReward Type=CPMI2026.04 | 0.765 | 34.6 | 6.34 | 7.85 | |
| MCReward Type=MC2026.04 | 0.759 | 27.7 | 1 | 1 | |
| PAVReward Type=PAV2026.04 | 0.757 | 36.6 | 0.85 | 1.12 |