CollabFlow：智能体协作的递归自我改进

AI 摘要

该论文提出 CollabFlow，一个面向 LLM 多智能体协作的递归自我改进（RSI）系统。它由可训练的 Collab-Director 构建完整智能体团队、冻结执行器运行团队，并用每轮结果重新训练 director。方法引入证据条件通信协议与 Collaborative Trajectory Balance 目标，在十二个数据集上优于所有基线并随轮次持续提升。

正文节选

CollabFlow: Recursive Self-Improvement of Agent Collaboration Abstract Recursive self-improvement (RSI) lets a system improve from its own outcomes; in LLM-based multi-agent systems, Agents refine one another within a task, and outcomes improve how they collaborate across tasks. However, existing multi-agent collaboration leaves this loop open: collaboration is pre-defined at the operator level, topology-only learning keeps verbatim exchange that propagates errors, and reward maximization on a s

发布时间：2026-10-01 12:00

抓取时间：2026-10-01 12:42

来源机构：arXiv