Tool Use on API-Bank L2 cleaned (test)
87.32F1 (API Matching)ToolPRM
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| ToolPRMBackbone=Hammer2.1-1.5B, Inference Strategy=ToolPRM2025.10 | 87.32 | 68.89 | 78.11 | |
| Hammer2.1-1.5B + Best of NBackbone=Hammer2.1-1.5B, Inference Strategy=Best of N2025.10 | 86.98 | 68.49 | 77.74 | |
| Hammer2.1-1.5B + MajorityBackbone=Hammer2.1-1.5B, Inference Strategy=Majority2025.10 | 85.71 | 68.57 | 77.14 | |
| Hammer2.1-1.5BBackbone=Hammer2.1-1.5B2025.10 | 85.09 | 67.82 | 76.46 | |
| Hammer2.1-1.5B + Token-level Beam SearchBackbone=Hammer2.1-1.5B, Inference Strategy=Token-level Beam Search2025.10 | 81.5 | 63.84 | 72.67 |