InterLV-Search：交错多模态代理搜索基准测试
InterLV-Search: Benchmarking Interleaved Multimodal Agentic Search
AI总结：
本文提出InterLV-Search基准测试，用于评估交错语言-视觉代理搜索，包含多层级任务，涵盖主动视觉证据搜索、受控离线交错多模态搜索和开放网页交错多模态搜索，揭示当前系统在视觉证据获取、搜索控制及多模态证据整合方面的挑战。
摘要：
现有多模态代理搜索基准测试评估多模态搜索和视觉浏览，但视觉证据要么局限于输入，要么被视为答案终点而非交错搜索轨迹的一部分。我们引入InterLV-Search，一个用于交错语言-视觉代理搜索的基准测试，其中文本和视觉证据被反复用于条件后续搜索。它包含2,061个示例，涵盖三个层级：主动视觉证据搜索、受控离线交错多模态搜索和开放网页交错多模态搜索。除了现有基准测试外，它还包含多模态多分支样本，涉及证据搜索过程中多个实体之间的比较。我们用自动化管道构建了层级1和2，用机器引导的人工监督开放网页管道构建了层级3。我们进一步提供InterLV-Agent用于标准化工具使用、轨迹记录和评估。在专有和开源多模态代理上的实验表明，当前系统距离解决交错多模态搜索仍有很大差距，最佳模型整体准确率低于50%，凸显了视觉证据获取、搜索控制和多模态证据整合的挑战。我们发布了基准测试数据和评估代码在https://github.com/hbhalpha/InterLV-Search-Bench
英文摘要：
Existing benchmarks for multimodal agentic search evaluate multimodal search and visual browsing, but visual evidence is either confined to the input or treated as an answer endpoint rather than part of an interleaved search trajectory. We introduce \textbf{InterLV-Search}, a benchmark for Interleaved Language-Vision Agentic Search, in which textual and visual evidence is repeatedly used to condition later search. It contains 2,061 examples across three levels: active visual evidence seeking, controlled offline interleaved multimodal search, and open-web interleaved multimodal search. Beyond existing benchmarks, it also includes multimodal multi-branch samples that involve comparison between multiple entities during the evidence search. We construct Level 1 and Level 2 with automated pipelines and Level 3 with a machine-led, human-supervised open-web pipeline. We further provide InterLV-Agent for standardized tool use, trajectory logging, and evaluation. Experiments on proprietary and open-source multimodal agents show that current systems remain far from solving interleaved multimodal search, with the best model below 50% overall accuracy, highlighting challenges in visual evidence seeking, search control, and multimodal evidence integration. We release the benchmark data and evaluation code at https://github.com/hbhalpha/InterLV-Search-Bench
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
