Tool-calling on BFCL v3 (test)
18Base AccuracyMemp
Evaluation Results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| MempModel=Llama3.2-3B-Instruct2026.06 | 18 | — | — | — | 10.75 | — | 8 | 9 | 8 | |
| KATE w/o PSModel=Llama3.2-3B-Instruct2026.06 | 18 | — | — | — | 12.75 | — | 14 | 11 | 8 | |
| KATE r PS-Con.Model=Llama3.2-3B-Instruct2026.06 | 14 | — | — | — | 11 | — | 15 | 9 | 6 | |
| KATEModel=Llama3.2-3B-Instruct2026.06 | 12 | — | — | — | 11.25 | — | 15 | 11 | 7 | |
| KATE w/o ExpModel=Llama3.2-3B-Instruct2026.06 | 7 | — | — | — | 4.25 | — | 3 | 3 | 4 | |
| FCModel=Llama3.2-3B-Instruct2026.06 | 2 | — | — | — | 2 | — | 3 | 1 | 2 |