Function Calling on BFCL Multiple v3
84.5AccuracyFree-form d=16
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Free-form d=16Budget=16, Model Size=7B, Decoding strategy=greedy2026.04 | 84.5 | — | 1 | |
| FR-CoTBudget=≤30, Model Size=7B, Decoding strategy=greedy2026.04 | 83 | 98.5 | 0 | |
| Free-form d=32 (default)Budget=32, Model Size=7B, Decoding strategy=greedy2026.04 | 82.5 | — | 2.5 | |
| Free-form d=16Budget=16, Model Size=1.5B, Decoding strategy=greedy2026.04 | 71.5 | — | 1 | |
| Free-form d=32 (default)Budget=32, Model Size=1.5B, Decoding strategy=greedy2026.04 | 64 | — | 2.5 | |
| FR-CoTBudget=≤30, Model Size=1.5B, Decoding strategy=greedy2026.04 | 64 | 99.5 | 0 | |
| Free-form d=0 (no CoT)Budget=0, Model Size=1.5B, Decoding strategy=greedy2026.04 | 44 | — | 3 | |
| Free-form d=0 (no CoT)Budget=0, Model Size=7B, Decoding strategy=greedy2026.04 | 40.5 | — | 3 |