Tool Calling on BFCL out-of-domain
26Exact Match (EM)TInR-U
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| TInR-U2026.04 | 26 | 35.83 | 50.12 | |
| ToolGen2026.04 | 22.01 | 30.91 | 48.24 | |
| ToolRetriever+Qwen3-8BRetriever=ToolRetriever, Reasoning=Qwen3-8B2026.04 | 17.45 | 36.42 | 29.27 | |
| ToolRetriever+ToolRLRetriever=ToolRetriever, Reasoning=ToolRL2026.04 | 16.63 | 37.35 | 28.45 | |
| TR-Feedback+ToolRLRetriever=TR-Feedback, Reasoning=ToolRL2026.04 | 16.39 | 35.71 | 29.63 | |
| BM25+ToolRLRetriever=BM25, Reasoning=ToolRL2026.04 | 16.1 | 32.55 | 26.46 | |
| Ada-embedding+ToolRLRetriever=Ada-embedding, Reasoning=ToolRL2026.04 | 16.04 | 36.82 | 28.92 | |
| ToolRetriever+Hammer2.1-7bRetriever=ToolRetriever, Reasoning=Hammer2.1-7b2026.04 | 12.06 | 28.57 | 22.72 | |
| ATU2026.04 | 11.59 | 25.29 | 33.37 | |
| ToolRetriever+xLAM-7b-rRetriever=ToolRetriever, Reasoning=xLAM-7b-r2026.04 | 10.13 | 27.17 | 21.55 |