Multi-turn Reasoning on All-benchmark Average
0.683Average Multi-turn Accuracy (multi@5)DRIFT-5turn
Evaluation results
| Method | Links | |
|---|---|---|
| DRIFT-5turnModel=Qwen2.5-7B-Instruct, Maximum budget=5 turns2026.05 | 0.683 | |
| UFO-5turnModel=Qwen2.5-7B-Instruct, Maximum budget=5 turns2026.05 | 0.679 | |
| SFT-5turnModel=Qwen2.5-7B-Instruct, Maximum budget=5 turns2026.05 | 0.656 | |
| BaseModel=Qwen2.5-7B-Instruct, Maximum budget=5 turns2026.05 | 0.648 |