作者位次 #9/21ACL 2026Poster· Natural Language Processing
B rowse C omp-Plus: A Fair and Disentangled Evaluation Benchmark for Deep Search Agents
Zijian Chen · Xueguang Ma · Shengyao Zhuang · Ping Nie · Kai Zou · Sahel Sharifymoghaddam · Andrew Liu · Joshua Green · Kshama Patel · Ruoxi Meng · Mingyi Su · Yanxi Li · Haoran Hong · Xinyu Shi · Xuye Liu · Hosna Oyarhoseini · Nandan Thakur · Crystina Zhang · Luyu Gao · Wenhu Chen · Jimmy Lin
摘要
Deep search agents that combine large language models with retrieval tools excel at complex, multi-hop queries. Yet, existing benchmarks such as BrowseComp rely on black-box web search APIs, facing key limitations. (1) Fairness: for agents, dynamic and opaque web APIs hinder reproducibility and f...