Function Calling on Berkeley Function Calling Leaderboard (BFCL)
68.92Overall SuccessTIER
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| TIERReward format=TIER2026.05 | 68.92 | 85.94 | 81.96 | 39.12 | 77.78 | 87.65 | |
| SimpleReward format=Simple2026.05 | 66.3 | 85 | 78.72 | 34 | 77.78 | 83.56 | |
| BaseReward format=Base, Model=Qwen3-8B2026.05 | 64.31 | 88.81 | 78.54 | 33.38 | 77.78 | 79.08 | |
| Supervised Fine TuningReward format=Supervised Fine Tuning2026.05 | 61.47 | 84.12 | 79.52 | 19.25 | 61.11 | 90.33 | |
| ToolRLReward format=ToolRL2026.05 | 37.27 | 84.1 | 43.98 | 0.38 | 94.44 | 0.81 |