智能体视频智能：用于高级视频探索与理解的灵活框架
Agentic Video Intelligence: A Flexible Framework for Advanced Video Exploration and Understanding
AI总结：
针对视频理解中VLMs单遍处理、证据重访受限及现有智能体方法依赖专有模型或需大量训练的问题，提出无需训练的AVI框架，通过三阶段推理、结构化知识库及开源模型集成，实现类人视频理解并在多基准上取得优效。
摘要：
视频理解不仅需要视觉识别，还需要复杂推理。尽管视觉-语言模型（VLMs）展现出令人印象深刻的能力，但它们通常以单遍方式处理视频，对证据重访和迭代优化的支持有限。近期涌现的基于智能体的方法虽能实现长程推理，但要么严重依赖昂贵的专有模型，要么需要大量智能体强化学习（RL）训练。为克服这些局限，我们提出智能体视频智能（AVI），这是一个灵活且无需训练的框架，可通过系统级设计与优化模拟人类视频理解。AVI引入三项关键创新：（1）受人类启发的三阶段推理过程（检索-感知-回顾），确保充分的全局探索与聚焦的局部分析；（2）通过实体图组织的结构化视频知识库，以及多粒度集成工具，构成智能体的交互环境；（3）开源模型集成，结合推理大语言模型（LLMs）与轻量级基础计算机视觉（CV）模型及VLM，消除对专有API或RL训练的依赖。在LVBench、VideoMME-Long、LongVideoBench和Charades-STA上的实验表明，AVI在实现竞争力性能的同时，提供了更优的可解释性。
英文摘要：
Video understanding requires not only visual recognition but also complex reasoning. While Vision-Language Models (VLMs) demonstrate impressive capabilities, they typically process videos largely in a single-pass manner with limited support for evidence revisit and iterative refinement. While recently emerging agent-based methods enable long-horizon reasoning, they either depend heavily on expensive proprietary models or require extensive agentic RL training. To overcome these limitations, we propose Agentic Video Intelligence (AVI), a flexible and training-free framework that can mirror human video comprehension through system-level design and optimization. AVI introduces three key innovations: (1) a human-inspired three-phase reasoning process (Retrieve-Perceive-Review) that ensures both sufficient global exploration and focused local analysis, (2) a structured video knowledge base organized through entity graphs, along with multi-granularity integrated tools, constituting the agent's interaction environment, and (3) an open-source model ensemble combining reasoning LLMs with lightweight base CV models and VLM, eliminating dependence on proprietary APIs or RL training. Experiments on LVBench, VideoMME-Long, LongVideoBench, and Charades-STA demonstrate that AVI achieves competitive performance while offering superior interpretability.
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。
