List your company

Tool Invocation Refusal on SABEval (D1)

59.46Tool Invocation Rate (TIR)Base
0.720815.970431.2246.4696Apr 13, 2026

Evaluation results

MethodLinks
BaseModel=Qwen3-14B2026.0459.46
BaseModel=Qwen3-4B2026.0457.85
PromptModel=ToolACE-2.52026.0453.66
BaseModel=Qwen3-8B2026.0451.61
BaseModel=ToolACE-2.52026.0451.12
PromptModel=Qwen3-4B2026.0442.2
PromptModel=Qwen3-8B2026.0434.29
PromptModel=Qwen3-14B2026.0433.41
BaseModel=Watt-Tool2026.0415.46
PromptModel=Watt-Tool2026.0415.37
CAAModel=Qwen3-4B2026.0414.29
CAAModel=Qwen3-14B2026.047.27
CAAModel=Qwen3-8B2026.046.49
CAAModel=ToolACE-2.52026.045.27
CAAModel=Watt-Tool2026.042.98