LensWalk：通过规划视频观看方式实现智能体式视频理解
LensWalk: Agentic Video Understanding by Planning How You See in Videos
把「看哪里、看多密」变成智能体可调的工具参数，视觉研究者可借鉴这种按需采样+工具编排的推理范式，用于长视频问答与证据定位，无需微调即可插拔到现有 VLM 配方上。
原文依据 the agent dynamically specifies, at each step, the temporal scope and sampling density of the video it observes
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论视频的稠密、时序特性给自动化分析带来了深刻挑战。尽管视觉语言模型能力强大，现有视频理解方法仍受制于推理与感知之间的固有脱节：它们依赖静态、预处理后的信息，无法随着理解过程的演进而主动从视频中寻找原始证据。为解决这一问题，我们提出 LensWalk，一个灵活的智能体框架，使大语言模型推理器能够主动控制自身的视觉观察。LensWalk 建立了一个紧密的推理—规划—观察循环，智能体在每一步动态指定其所观察视频的时间范围与采样密度。借助一组以这些规格为参数、基于视觉语言模型的多功能工具，智能体可以执行广泛扫描以寻找线索、聚焦特定片段以抽取事实，并拼接来自多个时刻的证据进行整体验证。这一设计支持渐进式、按需的证据收集，直接服务于智能体不断演进的思维链。无需任何模型微调，LensWalk 在多种模型配方上带来显著的即插即用性能提升，在 LVBench 和 Video-MME 等具有挑战性的长视频基准上将其准确率提升超过 5%。我们的分析表明，让智能体控制其观看方式，是解锁更准确、更鲁棒、更可解释的视频推理的关键。
英文原文摘要 1484 字符
The dense, temporal nature of video presents a profound challenge for automated analysis. Despite the use of powerful Vision-Language Models, prevailing methods for video understanding are limited by the inherent disconnect between reasoning and perception: they rely on static, pre-processed information and cannot actively seek raw evidence from video as their understanding evolves. To address this, we introduce LensWalk, a flexible agentic framework that empowers a Large Language Model reasoner to control its own visual observation actively. LensWalk establishes a tight reason-plan-observe loop where the agent dynamically specifies, at each step, the temporal scope and sampling density of the video it observes. Using a suite of versatile, Vision-Language Model based tools parameterized by these specifications, the agent can perform broad scans for cues, focus on specific segments for fact extraction, and stitch evidence from multiple moments for holistic verification. This design allows for progressive, on-demand evidence gathering that directly serves the agent's evolving chain of thought. Without requiring any model fine-tuning, LensWalk delivers substantial, plug-and-play performance gains on multiple model recipes, boosting their accuracy by over 5% on challenging long-video benchmarks like LVBench and Video-MME. Our analysis reveals that enabling an agent to control how it sees is key to unlocking more accurate, robust, and interpretable video reasoning.