Deep search requires LLM agents to decompose complex queries, search for evidence, and synthesize grounded answers, yet existing ReAct-style agents suffer fr…
机构:腾讯
来源:arXiv 2609.29444 | AI4Papers 论文推荐平台