Mathematical Reasoning on OlympiadBench (Pass@1 (%))
68.1Pass@1 AccuracyOLMo-32B-Instruct (Teacher)
Evaluation Results
| Method | Links | |
|---|---|---|
| OLMo-32B-Instruct (Teacher)Model Category=Off-the-shelf2026.06 | 68.1 | |
| MIPUModel=Qwen3-4B, Rollout Quantization=FP82026.06 | 67.86 | |
| LR-decayModel=Qwen3-4B, Rollout Quantization=FP82026.06 | 67.26 | |
| GRPO BaselineModel=Qwen3-4B, Rollout Quantization=FP82026.06 | 64.89 | |
| OmniOPDStudent Model=Qwen3-4B, Teacher Model=Gemini-2.5-flash, Method=OmniOPD2026.05 | 64.58 | |
| OmniOPDStudent Model=Qwen3-4B, Teacher Model=Claude-4.5-haiku, Method=OmniOPD2026.05 | 64.35 | |
| Claude-4.5-haikuStudent Model=-, Teacher Model=Claude-4.5-haiku, Method=Base2026.05 | 62.69 | |
| SFTStudent Model=Qwen3-4B, Teacher Model=Gemini-2.5-flash, Method=SFT2026.05 | 62.61 | |
| Token Zones + KL AnnealingModel Category=Proposed (SEAD components)2026.06 | 62.5 | |
| MISModel=Qwen3-4B, Rollout Quantization=FP82026.06 | 62.5 | |
| SEAD (full)Model Category=Proposed (SEAD components)2026.06 | 62.1 | |
| Gemini-2.5-flashStudent Model=-, Teacher Model=Gemini-2.5-flash, Method=Base2026.05 | 61.86 | |
| OPD (RKL k=1)Model Category=Baseline2026.06 | 60.1 | |
| Token Zones onlyModel Category=Proposed (SEAD components)2026.06 | 60.1 | |
| MIPUModel=Qwen3-1.7B, Rollout Quantization=FP82026.06 | 59.52 | |
| LR-decayModel=Qwen3-1.7B, Rollout Quantization=FP82026.06 | 58.93 | |
| GRPOModel Category=Baseline2026.06 | 58.8 | |
| OmniOPDStudent Model=Qwen3-1.7B, Teacher Model=Qwen3-32B, Method=OmniOPD2026.05 | 58.64 | |
| KL Annealing onlyModel Category=Proposed (SEAD components)2026.06 | 58.5 | |
| MISModel=Qwen3-1.7B, Rollout Quantization=FP82026.06 | 58.33 | |
| OLMo-7B-Instruct (Student)Model Category=Off-the-shelf2026.06 | 58.2 | |
| OPDStudent Model=Qwen3-1.7B, Teacher Model=Qwen3-32B, Method=OPD2026.05 | 58.08 | |
| OPSDModel Category=Baseline2026.06 | 57.8 | |
| SFTStudent Model=Qwen3-4B, Teacher Model=Claude-4.5-haiku, Method=SFT2026.05 | 57.71 | |
| GRPO BaselineModel=Qwen3-1.7B, Rollout Quantization=FP82026.06 | 56.55 | |
| Qwen3-30B-A3B-Inst.Student Model=-, Teacher Model=Qwen3-30B-A3B-Inst., Method=Base2026.05 | 56.38 | |
| OmniOPDStudent Model=Qwen3-4B, Teacher Model=Qwen3-30B-A3B-Inst., Method=OmniOPD2026.05 | 56.05 | |
| Qwen3-1.7B (GRPO)Student Model=Qwen3-1.7B, Teacher Model=-, Method=GRPO2026.05 | 55.67 | |
| Qwen3-32BStudent Model=-, Teacher Model=Qwen3-32B, Method=Base2026.05 | 52.93 | |
| SFTStudent Model=Qwen3-1.7B, Teacher Model=Qwen3-32B, Method=SFT2026.05 | 52.19 | |
| Qwen3-4B (GRPO)Student Model=Qwen3-4B, Teacher Model=-, Method=GRPO2026.05 | 52.19 | |
| OmniOPDStudent Model=Qwen3-4B, Teacher Model=Qwen3-32B, Method=OmniOPD2026.05 | 50.27 | |
| OPDStudent Model=Qwen3-4B, Teacher Model=Qwen3-30B-A3B-Inst., Method=OPD2026.05 | 50.14 | |
| OPDStudent Model=Qwen3-4B, Teacher Model=Qwen3-32B, Method=OPD2026.05 | 48.96 | |
| SFTStudent Model=Qwen3-4B, Teacher Model=Qwen3-32B, Method=SFT2026.05 | 47.19 | |
| SFTStudent Model=Qwen3-4B, Teacher Model=Qwen3-30B-A3B-Inst., Method=SFT2026.05 | 46.29 | |
| Qwen3-4BStudent Model=Qwen3-4B, Teacher Model=-, Method=Base2026.05 | 40.32 | |
| Qwen3-1.7BStudent Model=Qwen3-1.7B, Teacher Model=-, Method=Base2026.05 | 37.9 |