Mathematical Reasoning on AIME26 (accuracy)
44.06AIME26 AccuracyTOP-D (Ours)
Evaluation Results
| Method | Links | |
|---|---|---|
| TOP-D (Ours)Student Model=Qwen3-8B-Base, Teacher Model=Qwen3-30B-A3B-Instruct-2507, Paradigm=Distillation, Evaluation Protocol=avg@322026.07 | 44.06 | |
| DAPOStudent Model=Qwen3-8B-Base, Teacher Model=Qwen3-30B-A3B-Instruct-2507, Paradigm=RLVR, Evaluation Protocol=avg@322026.07 | 32.29 | |
| OPDStudent Model=Qwen3-8B-Base, Teacher Model=Qwen3-30B-A3B-Instruct-2507, Paradigm=Distillation, Evaluation Protocol=avg@322026.07 | 25.42 | |
| GRPOStudent Model=Qwen3-8B-Base, Teacher Model=Qwen3-30B-A3B-Instruct-2507, Paradigm=RLVR, Evaluation Protocol=avg@322026.07 | 21.67 | |
| BaseStudent Model=Qwen3-8B-Base, Paradigm=-, Evaluation Protocol=avg@322026.07 | 6.15 |