List your company

Agentic Search on Combined General AI Assistant, WebWalkerQA, BrowseComp, XBench

11.8Overall ScoreReAct + GraphPO
5.4567.1038.7510.397Jun 17, 2026

Evaluation results

MethodLinks
ReAct + GraphPOBase Model=Qwen2.5-7B, Agent Framework=ReAct2026.0611.8
ReAct + Tree-GRPOBase Model=Qwen2.5-7B, Agent Framework=ReAct2026.0610.3
ReAct + DAPOBase Model=Qwen2.5-7B, Agent Framework=ReAct2026.069.8
ReAct + GRPOBase Model=Qwen2.5-7B, Agent Framework=ReAct2026.069.4
ReActBase Model=Qwen2.5-7B, Agent Framework=ReAct2026.065.7