Function-name selection on BFCL multiple live (300 tasks model)
97.7AccuracyQwen2.5-1.5B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Qwen2.5-1.5Bsetting=Oracle δ+82026.06 | 97.7 | — | |
| Llama-3.1-8Bsetting=Oracle δ+82026.06 | 96 | — | |
| Pooled (3)setting=Oracle δ+82026.06 | 95.9 | — | |
| Qwen2.5-7Bsetting=Oracle δ+82026.06 | 94 | — | |
| Llama-3.1-8Bsetting=Gold-free S22026.06 | 91.7 | 11.3 | |
| Qwen3-8Bsetting=Gold-free S22026.06 | 91.3 | 7.7 | |
| Qwen2.5-7Bsetting=Gold-free S22026.06 | 90 | 17.7 | |
| Pooled (3)setting=Gold-free S22026.06 | 89.9 | 11.9 | |
| Pooled (5)setting=Gold-free S22026.06 | 89.1 | 11.2 | |
| Qwen2.5-1.5Bsetting=Gold-free S22026.06 | 88 | 6.7 | |
| xLAM-2-8Bsetting=Gold-free S22026.06 | 84.7 | 12.7 | |
| Qwen3-8Bsetting=Base2026.06 | 83.7 | — | |
| Qwen2.5-1.5Bsetting=Base2026.06 | 81.3 | — | |
| Llama-3.1-8Bsetting=Base2026.06 | 80.3 | — | |
| Pooled (3)setting=Base2026.06 | 78 | — | |
| Pooled (5)setting=Base2026.06 | 77.9 | — | |
| Qwen2.5-7Bsetting=Base2026.06 | 72.3 | — | |
| xLAM-2-8Bsetting=Base2026.06 | 72 | — |