List your company

Tool Invocation Refusal on SABEval (D4)

91.02Tool Invocation Rate (TIR)Base
7.92429.49751.0772.643Apr 13, 2026

Evaluation results

MethodLinks
BaseModel=Qwen3-14B2026.0491.02
BaseModel=Qwen3-4B2026.0489.76
BaseModel=Qwen3-8B2026.0482.44
PromptModel=Qwen3-4B2026.0475.95
BaseModel=ToolACE-2.52026.0468
PromptModel=ToolACE-2.52026.0467.71
PromptModel=Qwen3-8B2026.0466.2
PromptModel=Qwen3-14B2026.0465.66
BaseModel=Watt-Tool2026.0439.51
PromptModel=Watt-Tool2026.0438.05
CAAModel=Qwen3-4B2026.0436
CAAModel=Qwen3-14B2026.0421.22
CAAModel=Watt-Tool2026.0412.68
CAAModel=ToolACE-2.52026.0411.46
CAAModel=Qwen3-8B2026.0411.12