Per-step Tool Selection Prediction on τ-bench airline multi-step
100Success RateQwen2.5-3B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Qwen2.5-3Bstatus=ceiling2026.06 | 100 | 1 | 99.8 | |
| Llama-3.1-8Bstatus=0 fails2026.06 | 100 | — | — | |
| Qwen2.5-7Bstatus=noise2026.06 | 95 | 0.387 | 23.6 | |
| Qwen2.5-0.5Bstatus=inform.2026.06 | 55 | 0.985 | 65.3 | |
| Qwen2.5-1.5Bstatus=inform.2026.06 | 49 | 0.998 | 97 |