LLM代理框架实现大型解空间的持续改进与并行自主探索
原标题：Continuous Improvement and Parallel Autonomous Exploration: An LLM-Agent Framework for Searching Large Solution Spaces
AI 摘要
该研究提出一个框架，为LLM代理提供两种自主搜索大型解空间的机制：基于留出数据评分的排行榜作为奖励信号驱动单代理持续改进，以及多代理并行自主探索。在电商产品目录匹配任务上，单代理达到47.8-57.4%的合格覆盖率，五代理达到62.8-69.4%，优于33.3%的基线。该框架贡献在于持续改进奖励循环和全自主并行探索机制。
正文节选
Computer Science > Multiagent Systems Title:Continuous Improvement and Parallel Autonomous Exploration: An LLM-Agent Framework for Searching Large Solution Spaces View PDF HTML (experimental) Abstract:We present a framework that gives LLM agents two mechanisms for searching large solution spaces autonomously. First, a leaderboard scored on held-out data acts as a reward signal that drives each agent to refine its solutions over repeated submissions, a loop that operates even with a s