Math Reasoning on AIME 2024 (Acc, Tok, AES)
0.627AccuracyExOPD
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ExOPDDistillation Setting=Single-Teacher Distillation2026.02 | 0.627 | — | — | |
| ExOPDDistillation Setting=Multi-Teacher Distillation2026.02 | 0.61 | — | — | |
| OPDDistillation Setting=Single-Teacher Distillation2026.02 | 0.607 | — | — | |
| OPDDistillation Setting=Multi-Teacher Distillation2026.02 | 0.606 | — | — | |
| ExPODistillation Setting=Single-Teacher Distillation2026.02 | 0.587 | — | — | |
| TLMREModel Size=7B, Learning Paradigm=outcome-supervised RL2026.02 | 0.585 | 7,462 | 0.61 | |
| SFTDistillation Setting=Multi-Teacher Distillation2026.02 | 0.585 | — | — | |
| Laser-DModel Size=7B, Learning Paradigm=outcome-supervised RL2026.02 | 0.583 | 5,379 | 0.79 | |
| TeacherDistillation Setting=Baseline (Teacher)2026.02 | 0.58 | — | — | |
| DIETModel Size=7B, Learning Paradigm=outcome-supervised RL2026.02 | 0.579 | 10,124 | 0.32 | |
| ExPODistillation Setting=Multi-Teacher Distillation2026.02 | 0.575 | — | — | |
| ATTNPOModel Size=7B, Learning Paradigm=process-supervised RL2026.02 | 0.572 | 5,214 | 0.75 | |
| S-GRPOModel Size=7B, Learning Paradigm=process-supervised RL2026.02 | 0.56 | 7,377 | 0.47 | |
| Laser-DEModel Size=7B, Learning Paradigm=outcome-supervised RL2026.02 | 0.558 | 4,969 | 0.69 | |
| AdaptThinkModel Size=7B, Learning Paradigm=adaptive-mode2026.02 | 0.556 | 8,599 | 0.33 | |
| AutoThinkModel Size=7B, Learning Paradigm=adaptive-mode2026.02 | 0.548 | 8,051 | 0.34 | |
| DS-7BModel Size=7B2026.02 | 0.529 | 10,490 | 0 | |
| ACPOModel Size=7B, Learning Paradigm=outcome-supervised RL2026.02 | 0.528 | 4,520 | 0.56 | |
| DEPOModel Size=7B, Learning Paradigm=process-supervised RL2026.02 | 0.527 | 6,580 | 0.35 | |
| VSRM-R++Model Size=7B, Learning Paradigm=process-supervised RL2026.02 | 0.522 | 6,773 | 0.29 | |
| DECSModel Size=7B, Learning Paradigm=process-supervised RL2026.02 | 0.513 | 5,277 | 0.35 | |
| ATTNPOModel Size=1.5B, Learning Paradigm=process-supervised RL2026.02 | 0.377 | 5,713 | 1.56 | |
| TLMREModel Size=1.5B, Learning Paradigm=outcome-supervised RL2026.02 | 0.373 | 7,024 | 1.41 | |
| Laser-DEModel Size=1.5B, Learning Paradigm=outcome-supervised RL2026.02 | 0.35 | 5,789 | 1.27 | |
| AutoThinkModel Size=1.5B, Learning Paradigm=adaptive-mode2026.02 | 0.346 | 9,514 | 0.92 | |
| Laser-DModel Size=1.5B, Learning Paradigm=outcome-supervised RL2026.02 | 0.342 | 5,750 | 1.19 | |
| DIETModel Size=1.5B, Learning Paradigm=outcome-supervised RL2026.02 | 0.318 | 10,578 | 0.54 | |
| DECSModel Size=1.5B, Learning Paradigm=process-supervised RL2026.02 | 0.312 | 5,550 | 0.89 | |
| AdaptThinkModel Size=1.5B, Learning Paradigm=adaptive-mode2026.02 | 0.31 | 6,670 | 0.77 | |
| DEPOModel Size=1.5B, Learning Paradigm=process-supervised RL2026.02 | 0.308 | 7,732 | 0.66 | |
| ACPOModel Size=1.5B, Learning Paradigm=outcome-supervised RL2026.02 | 0.3 | 6,670 | 0.66 | |
| VSRM-R++Model Size=1.5B, Learning Paradigm=process-supervised RL2026.02 | 0.295 | 6,958 | 0.59 | |
| DS-1.5BModel Size=1.5B2026.02 | 0.281 | 12,339 | 0 | |
| ThinkPruneModel Size=1.5B, Learning Paradigm=outcome-supervised RL2026.02 | 0.271 | 5,631 | 0.37 | |
| LC-R1Model Size=1.5B, Learning Paradigm=process-supervised RL2026.02 | 0.236 | 7,122 | -0.37 | |
| LC-R1Model Size=7B, Learning Paradigm=process-supervised RL2026.02 | 0.236 | 6,904 | -2.42 | |
| StudentDistillation Setting=Baseline (Student)2026.02 | 0.215 | — | — |