BAPO: Boundary-Aware Policy Optimization for Reliable Agentic Search
BAPO:面向可靠代理搜索的边界感知策略优化
机构 * Institute of Artificial Intelligence, Xiamen University(厦门大学人工智能研究院) ; Meituan Inc.(美团公司) ; School of Informatics, Xiamen University(厦门大学信息学院) ; Westlake University(西湖大学) ; Jilin University(吉林大学) ; Key Laboratory of Digital Protection and Intelligent Processing of Intangible Cultural Heritage of Fujian and Taiwan (Xiamen University), Ministry of Culture and Tourism, China(福建省和台湾非物质文化遗产数字化保护与智能处理重点实验室(厦门大学),中华人民共和国文旅部,中国)
AI总结 BAPO通过引入边界感知奖励和自适应奖励调节器,提升代理搜索的可靠性,避免过度依赖IDK响应,实验表明其显著增强可靠性。
Comments ACL 2026 Conference. Code is available at https://github.com/Liushiyu-0709/BAPO-Reliable-Search