List your company

Tool Invocation Refusal on SABEval (D3)

85.56Tool Invocation Rate (TIR)Base
3.732824.976446.2267.4636Apr 13, 2026

Evaluation results

MethodLinks
BaseModel=Qwen3-4B2026.0485.56
BaseModel=Qwen3-14B2026.0483.9
BaseModel=Qwen3-8B2026.0475.56
PromptModel=Qwen3-4B2026.0469.71
PromptModel=ToolACE-2.52026.0467.71
BaseModel=ToolACE-2.52026.0467.41
PromptModel=Qwen3-8B2026.0457.66
PromptModel=Qwen3-14B2026.0456.78
BaseModel=Watt-Tool2026.0431.51
PromptModel=Watt-Tool2026.0429.61
CAAModel=Qwen3-4B2026.0427.76
CAAModel=Qwen3-14B2026.0415.51
CAAModel=Qwen3-8B2026.048.98
CAAModel=Watt-Tool2026.048.1
CAAModel=ToolACE-2.52026.046.88