Math Reasoning on AMC 2023 (Acc, Tok, AES)
90.5AccuracyDEPO
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| DEPOModel Size=7B, Learning Paradigm=process-supervised RL2026.02 | 90.5 | 3,215 | 0.58 | |
| Laser-DModel Size=7B, Learning Paradigm=outcome-supervised RL2026.02 | 90 | 2,694 | 0.65 | |
| TLMREModel Size=7B, Learning Paradigm=outcome-supervised RL2026.02 | 89.9 | 2,967 | 0.6 | |
| ATTNPOModel Size=7B, Learning Paradigm=process-supervised RL2026.02 | 89.6 | 2,303 | 0.7 | |
| Laser-DEModel Size=7B, Learning Paradigm=outcome-supervised RL2026.02 | 89.1 | 2,612 | 0.63 | |
| DECSModel Size=7B, Learning Paradigm=process-supervised RL2026.02 | 89 | 2,772 | 0.6 | |
| AdaptThinkModel Size=7B, Learning Paradigm=adaptive-mode2026.02 | 87.5 | 3,287 | 0.46 | |
| S-GRPOModel Size=7B, Learning Paradigm=process-supervised RL2026.02 | 87.5 | 3,494 | 0.43 | |
| DS-7BModel Size=7B2026.02 | 87.1 | 5,977 | 0 | |
| AutoThinkModel Size=7B, Learning Paradigm=adaptive-mode2026.02 | 83.3 | 4,645 | 0 | |
| DIETModel Size=7B, Learning Paradigm=outcome-supervised RL2026.02 | 82.6 | 6,075 | -0.27 | |
| VSRM-R++Model Size=7B, Learning Paradigm=process-supervised RL2026.02 | 80.9 | 3,704 | 0.02 | |
| TLMREModel Size=1.5B, Learning Paradigm=outcome-supervised RL2026.02 | 79.1 | 3,349 | 1.18 | |
| LC-R1Model Size=7B, Learning Paradigm=process-supervised RL2026.02 | 79.1 | 3,453 | -0.04 | |
| ATTNPOModel Size=1.5B, Learning Paradigm=process-supervised RL2026.02 | 77.8 | 2,638 | 1.21 | |
| DECSModel Size=1.5B, Learning Paradigm=process-supervised RL2026.02 | 75.4 | 2,988 | 1.05 | |
| Laser-DModel Size=1.5B, Learning Paradigm=outcome-supervised RL2026.02 | 75.3 | 2,981 | 1.05 | |
| DEPOModel Size=1.5B, Learning Paradigm=process-supervised RL2026.02 | 74.2 | 4,388 | 0.82 | |
| Laser-DEModel Size=1.5B, Learning Paradigm=outcome-supervised RL2026.02 | 73.3 | 3,080 | 0.95 | |
| ThinkPruneModel Size=1.5B, Learning Paradigm=outcome-supervised RL2026.02 | 73.2 | 3,039 | 0.95 | |
| AutoThinkModel Size=1.5B, Learning Paradigm=adaptive-mode2026.02 | 67 | 5,059 | 0.41 | |
| AdaptThinkModel Size=1.5B, Learning Paradigm=adaptive-mode2026.02 | 67 | 3,510 | 0.61 | |
| DS-1.5BModel Size=1.5B2026.02 | 65.9 | 7,893 | 0 | |
| DIETModel Size=1.5B, Learning Paradigm=outcome-supervised RL2026.02 | 65.4 | 6,425 | 0.15 | |
| VSRM-R++Model Size=1.5B, Learning Paradigm=process-supervised RL2026.02 | 64.7 | 4,119 | 0.39 | |
| LC-R1Model Size=1.5B, Learning Paradigm=process-supervised RL2026.02 | 61.7 | 3,947 | 0.18 |