arXiv cs.AI· Yuyang Wu, Yang Liu, Zequn Xie, Junjie Wang, Dan Yang, Zehao Tan, Ding-Chu Zhang, Junming Yang, Yue Shen, Jian Wang, Jinjie Gu·· 3 小时前AI 评分47
LEAD-RL：从委派到协调，用贡献感知信用分配训练高效多智能体深度研究
From Delegation to Coordination: Learning Efficient Multi-Agent Deep Research
AI 导读
论文提出 LEAD-RL，一个主智能体与子智能体联合训练框架，通过贡献感知的信用分配改进多智能体深度研究系统，最终答案正确性仍是唯一团队奖励，不使用 token、步数或加速奖励。在两个模型家族、五个基准上验证：BrowseComp 上较共享团队奖励的联合训练准确率提升 2.49 个百分点，关键路径成本降低 24.5%；验证阶段准确率提升 6.78 个百分点且关键路径成本基本不变。
正文 · 原文
Authors:Yuyang Wu, Yang Liu, Zequn Xie, Junjie Wang, Dan Yang, Zehao Tan, Ding-Chu Zhang, Junming Yang, Yue Shen, Jian Wang, Jinjie Gu
Abstract:Multi-agent deep-research systems use delegation to parallelize search and keep long tool traces out of the main agent's context. Yet delegation helps only when task allocation and subtask execution are coordinated: the main agent must issue useful subtasks, while subagents must return reliable evidence. Existing training either optimizes only the main agent or broadcasts the same team advantage to all agents, which can reward misleading subagent reports and penalize useful ones. We introduce LEAD-RL (Learning Effective Agent Delegation with Reinforcement Learning), a joint main-agent-subagent training framework with contribution-aware credit assignment. A training-time rubric evaluates the research utility of each subagent report and gates its inherited team advantage, while final-answer correctness remains the only team reward; no token, step, or speedup reward is used. Across five benchmarks and two model families, LEAD-RL improves answer quality. On BrowseComp, it improves accuracy by 2.49 percentage points and reduces critical-path cost by 24.5% over joint training with shared team credit. During validation, accuracy increases by 6.78 points at approximately constant critical-path cost. These results show that contribution-aware credit assignment improves the quality-efficiency trade-off of multi-agent deep research.
| Comments: | 16 pages, 3 figures, 3 tables |
| Subjects: | Multiagent Systems (cs.MA); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.17041 [cs.MA] |
| (or arXiv:2611.17041v1 [cs.MA] for this version) | |
| https://doi.org/10.48550/arXiv.2611.17041 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Yang Liu [view email]
[v1]
Thu, 12 Nov 2026 16:26:27 UTC (640 KB)
来源：arXiv cs.AI · arxiv.org