Scientific Reasoning on GPQA D (Pass@1 accuracy)
43.2Pass@1 AccuracyTeacher-OPPO
Evaluation results
| Method | Links | |
|---|---|---|
| Teacher-OPPOBase Model=Qwen3-4B-Instruct-2507, Optimization Framework=DAPO2026.05 | 43.2 | |
| Teacher-OPPOBase Model=Qwen3-4B-Instruct-2507, Optimization Framework=GRPO2026.05 | 42.5 | |
| Self-OPPOBase Model=Qwen3-4B-Instruct-2507, Optimization Framework=DAPO2026.05 | 41.8 | |
| Self-OPPOBase Model=Qwen3-4B-Instruct-2507, Optimization Framework=GRPO2026.05 | 41.6 | |
| DAPOBase Model=Qwen3-4B-Instruct-25072026.05 | 40.3 | |
| SDPOBase Model=Qwen3-4B-Instruct-25072026.05 | 40 | |
| Dr.GRPOBase Model=Qwen3-4B-Instruct-25072026.05 | 39.4 | |
| GRPOBase Model=Qwen3-4B-Instruct-25072026.05 | 39.2 | |
| SFTBase Model=Qwen3-4B-Instruct-25072026.05 | 37.8 | |
| Teacher-OPPOBase Model=Phi-4-mini-instruct, Optimization Framework=DAPO2026.05 | 36.4 | |
| Teacher-OPPOBase Model=Phi-4-mini-instruct, Optimization Framework=GRPO2026.05 | 35.8 | |
| Self-OPPOBase Model=Phi-4-mini-instruct, Optimization Framework=DAPO2026.05 | 35 | |
| Self-OPPOBase Model=Phi-4-mini-instruct, Optimization Framework=GRPO2026.05 | 34.5 | |
| DAPOBase Model=Phi-4-mini-instruct2026.05 | 33.5 | |
| SDPOBase Model=Phi-4-mini-instruct2026.05 | 33 | |
| Dr.GRPOBase Model=Phi-4-mini-instruct2026.05 | 32.6 | |
| GRPOBase Model=Phi-4-mini-instruct2026.05 | 32.4 | |
| SFTBase Model=Phi-4-mini-instruct2026.05 | 30.8 |