Mathematical Reasoning on AIME 2025 (ave@32)
68.8Accuracy (ave@32)Qwen3-4B + Direct-OPD
Evaluation Results
| Method | Links | |
|---|---|---|
| Qwen3-4B + Direct-OPDTransfer source=JustRL policy-shift2026.07 | 68.8 | |
| Qwen3-4BTransfer source=JustRL policy-shift2026.07 | 65.6 | |
| Teacher RLTransfer source=QuestA policy-shift2026.07 | 62.29 | |
| Teacher refTransfer source=QuestA policy-shift2026.07 | 49.5 | |
| R1-Distill-7B + Direct-OPDTransfer source=JustRL policy-shift2026.07 | 48.8 | |
| Qwen3-1.7B + Direct-OPDTransfer source=JustRL policy-shift2026.07 | 46.3 | |
| R1-Distill-7B + Direct-OPDTransfer source=QuestA policy-shift2026.07 | 44 | |
| Qwen3-1.7B + Direct-OPDTransfer source=QuestA policy-shift2026.07 | 43.1 | |
| R1-Distill-7BTransfer source=JustRL policy-shift2026.07 | 40.5 | |
| R1-Distill-7BTransfer source=QuestA policy-shift2026.07 | 39.5 | |
| Teacher RLTransfer source=JustRL policy-shift2026.07 | 37.5 | |
| Qwen3-1.7BTransfer source=JustRL policy-shift2026.07 | 36.8 | |
| Qwen3-1.7BTransfer source=QuestA policy-shift2026.07 | 36.8 | |
| Teacher refTransfer source=JustRL policy-shift2026.07 | 24 |