Math Reasoning on Harvard-MIT Mathematics Tournament 2025 November
20.31Avg@32SG-OPD
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SG-OPDSetting=Strong-to-weak, Teacher=Qwen3-4B-Non-Thinking-RL-Math, Student=Qwen3-1.7B, Training steps=5002026.06 | 20.31 | 50 | |
| OPDSetting=Strong-to-weak, Teacher=Qwen3-4B-Non-Thinking-RL-Math, Student=Qwen3-1.7B, Training steps=5002026.06 | 19.79 | 46.67 | |
| SeqKDSetting=Strong-to-weak, Teacher=Qwen3-4B-Non-Thinking-RL-Math, Student=Qwen3-1.7B, Training steps=5002026.06 | 18.96 | 33.33 | |
| GKDSetting=Strong-to-weak, Teacher=Qwen3-4B-Non-Thinking-RL-Math, Student=Qwen3-1.7B, Training steps=5002026.06 | 18.96 | 46.67 | |
| SFTSetting=Strong-to-weak, Teacher=Qwen3-4B-Non-Thinking-RL-Math, Student=Qwen3-1.7B, Training steps=5002026.06 | 18.33 | 31.41 | |
| ExOPDSetting=Strong-to-weak, Teacher=Qwen3-4B-Non-Thinking-RL-Math, Student=Qwen3-1.7B, Training steps=5002026.06 | 17.81 | 43.33 |