Tool Invocation Refusal on SABEval (D4)
91.02Tool Invocation Rate (TIR)Base
Evaluation Results
| Method | Links | |
|---|---|---|
| BaseModel=Qwen3-14B2026.04 | 91.02 | |
| BaseModel=Qwen3-4B2026.04 | 89.76 | |
| BaseModel=Qwen3-8B2026.04 | 82.44 | |
| PromptModel=Qwen3-4B2026.04 | 75.95 | |
| BaseModel=ToolACE-2.52026.04 | 68 | |
| PromptModel=ToolACE-2.52026.04 | 67.71 | |
| PromptModel=Qwen3-8B2026.04 | 66.2 | |
| PromptModel=Qwen3-14B2026.04 | 65.66 | |
| BaseModel=Watt-Tool2026.04 | 39.51 | |
| PromptModel=Watt-Tool2026.04 | 38.05 | |
| CAAModel=Qwen3-4B2026.04 | 36 | |
| CAAModel=Qwen3-14B2026.04 | 21.22 | |
| CAAModel=Watt-Tool2026.04 | 12.68 | |
| CAAModel=ToolACE-2.52026.04 | 11.46 | |
| CAAModel=Qwen3-8B2026.04 | 11.12 |