List your company

Tool-use Task Completion on ACEBench agent subset Multi-Turn

84.4Success RateReAct + EvoSOP
45.19255.37165.5575.729Jul 8, 2026

Evaluation results

MethodLinks
ReAct + EvoSOPBackbone=Qwen-Max2026.0784.4
DFSDT + EvoSOPBackbone=Gemini-3-FP2026.0783.3
ReAct + EvoSOPBackbone=Gemini-3-FP2026.0779.4
ReAct + EvoSOPBackbone=GPT-4o2026.0779.4
DFSDT + DRAFTBackbone=Qwen-Max2026.0778.3
DFSDT + EvoSOPBackbone=Qwen-Max2026.0777.9
DFSDT + EvoSOPBackbone=GPT-4o2026.0774.4
ReActBackbone=GPT-4o2026.0772.2
DFSDT + DRAFTBackbone=Gemini-3-FP2026.0771.1
DFSDTBackbone=GPT-4o2026.0769.4
DFSDT + DRAFTBackbone=GPT-4o2026.0766.1
ReAct + ASIBackbone=Qwen-Max2026.0765.6
ReAct + ASIBackbone=GPT-4o2026.0757.2
ReAct + ASIBackbone=Gemini-3-FP2026.0746.7