LongVidSearch: 一种多跳证据检索规划的代理基准
LongVidSearch: An Agentic Benchmark for Multi-hop Evidence Retrieval Planning in Long Videos
AI总结：
LongVidSearch 是一个评估长视频中多跳证据检索规划的基准，通过标准化接口强制多跳检索，包含3000个问题，涵盖四种推理类别，测试代理在相同访问条件下检索规划的准确性与效率。
摘要：
长视频问答（Long-Video QA）日益依赖代理工具使用来从长视频中检索证据。在现实场景中，这一过程通常需要多跳检索，即代理必须迭代收集多个不连续的证据片段。然而，现有长视频基准大多静态：它们很少强制严格的多跳检索，通常缺乏标准化的证据访问接口，使得难以区分检索规划失败与答案生成失败。为解决这一差距，我们引入LongVidSearch，一个评估长视频中代理多跳证据检索规划的基准，在标准化访问约束下。LongVidSearch强制检索必要性：一个Hop-k问题需要恰好k个必要证据片段，移除任何一个片段会使问题无法解决。该基准包含3000个问题，覆盖447个长视频（平均长度26分钟），涵盖四种推理类别：状态突变、因果推理、全局摘要和视觉跟踪，具有2跳、3跳和4跳的证据要求。为确保公平和受控的评估，所有代理通过统一的工具接口与LongVidSearch交互，固定检索后端并隔离代理制定查询和规划迭代检索的能力。除了答案准确性外，我们测量工具调用成本以分析在相同访问条件下准确性与效率的权衡。我们使用三种评委多数投票评估视频代理风格QA代理，使用多个基础LLM。GPT-5达到最高准确性（42.43），优于Gemini 3 Pro（30.97）和GPT-4o（19.20），但仍低于50%，凸显多跳检索规划的难度。有了黄金证据片段，性能变得接近完美，确认检索规划是主要瓶颈。
英文摘要：
Long video question answering (Long-Video QA) increasingly relies on agentic tool use to retrieve evidence from long videos. In realistic settings, this process often requires multi-hop retrieval, where agents must iteratively gather multiple discontinuous evidence clips. However, existing long-video benchmarks are largely static: they rarely enforce strict multi-hop retrieval and typically lack a standardized evidence-access interface, making it difficult to separate failures in retrieval planning from those in answer generation. To address this gap, we introduce LongVidSearch, a benchmark for evaluating agentic multi-hop evidence retrieval planning in long videos under standardized access constraints. LongVidSearch enforces retrieval necessity: a Hop-k question requires exactly k necessary evidence clips, and removing any single clip renders the question unsolvable. The benchmark contains 3,000 questions over 447 long videos (average length 26 minutes), covering four reasoning categories: State Mutation, Causal Inference, Global Summary, and Visual Tracking, with 2-hop, 3-hop, and 4-hop evidence requirements. To ensure fair and controlled evaluation, all agents interact with LongVidSearch through a unified tool interface, which fixes the retrieval backend and isolates the agent's ability to formulate queries and plan iterative retrieval. In addition to answer accuracy, we measure tool-call cost to analyze the accuracy-efficiency trade-off under identical access conditions. We evaluate VideoAgent-style QA agents with multiple backbone LLMs using three-judge majority voting. GPT-5 achieves the highest accuracy (42.43), outperforming Gemini 3 Pro (30.97) and GPT-4o (19.20), yet remaining below 50 %, highlighting the difficulty of multi-hop retrieval planning. With gold evidence clips, performance becomes near-perfect, confirming retrieval planning as the primary bottleneck.
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。
