Multi-turn Conversation on Multi-turn (Mt.)
55.4Mt. ScoreICPO
Evaluation Results
| Method | Links | |
|---|---|---|
| ICPOBase Model=Qwen2.5-7B-Instruct2026.01 | 55.4 | |
| Standard RLVRBase Model=Qwen2.5-7B-Instruct2026.01 | 47.2 | |
| Qwen2.5-7B-InstructModel=Qwen2.5-7B-Instruct2026.01 | 35.4 | |
| ICPOBase Model=Qwen2.5-1.5B-Instruct2026.01 | 32.8 | |
| Standard RLVRBase Model=Qwen2.5-1.5B-Instruct2026.01 | 26.1 | |
| Qwen2.5-1.5B-InstructModel=Qwen2.5-1.5B-Instruct2026.01 | 17 |