Tool Selection on ToolBench G3 (test)
25.8Recall@1L_PORTS
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| L_PORTSEncoder=BGE, Best LLM=CODESTRAL-22B-v0.12026.07 | 25.8 | 36.05 | 43.35 | 25.8 | 35.5 | 42.2 | 8.71 | 9.5 | |
| L_replugEncoder=BGE, Best LLM=LLAMA3-8B2026.07 | 22.56 | 33.74 | 40.51 | 22.56 | 33 | 36.65 | 5.92 | 5.73 | |
| L_PORTSEncoder=RoBERTa, Best LLM=CODESTRAL-22B-v0.12026.07 | 18.23 | 21.1 | 25.6 | 18.23 | 20.28 | 24.31 | 20.63 | 19.94 | |
| L_replugEncoder=RoBERTa, Best LLM=CODESTRAL-22B-v0.12026.07 | 12.56 | 20.11 | 24.8 | 12.56 | 19.67 | 22.19 | 18.14 | 17.14 |