arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2510.15440cs.CVcs.AI
收藏

少选多思:优先证据纯度以提升视频推理

Select Less, Reason More: Prioritizing Evidence Purity for Video Reasoning

CASIA(中国科学院自动化研究所);UCAS(中国科学院自动化研究所);ZGCA(中钢集团自动化研究所);NTU(国立台湾大学)

Xuchen Li, Xuzhao Li, Shiyu Hu, Kaiqi Huang

AI总结:

针对长视频推理中帧采样信息稀释和证据不纯的问题,提出证据感知强化学习(EARL)框架,动态选择关键帧并局部重采样,在五个基准上达到开源视频大语言模型最先进水平。

摘要:

长视频推理仍然是视频大语言模型(Video LLMs)面临的主要挑战,因为静态均匀帧采样会导致信息稀释并掩盖关键证据。此外,现有的像素空间视频推理智能体,虽然被设计为主动与视频交互以获取新的视觉信息,但由于缺乏严格的奖励机制来强制证据纯度,且无法在预采样帧之外进行时间信息补充,其性能仍然次优。为解决这一关键空白,我们提出了一种基于核心哲学“少选多思”的新型证据优先自适应框架。我们的核心贡献是证据感知强化学习(EARL)框架,该框架将模型转变为主动的证据询问者。EARL被精确设计用于动态选择最相关的帧,并且关键的是,在选定的关键帧周围进行局部重采样,以获取细粒度的时间细节。在五个具有挑战性的视频推理基准上的大量实验表明,我们经过EARL训练的模型在开源视频大语言模型中达到了新的最先进水平,同时学习了一种有效且高纯度的视觉证据选择策略。令人印象深刻的是,我们的7B模型在LongVideoBench上达到59.8%,在MVBench上达到69.0%,在VideoMME上达到64.9%。这些结果凸显了优先考虑证据纯度的重要性以及我们框架的有效性。

英文摘要:

Long-form video reasoning remains a major challenge for Video Large Language Models (Video LLMs), as static uniform frame sampling leads to information dilution and obscures critical evidence. Furthermore, existing pixel-space video reasoning agents, which are designed to actively interact with the video to acquire new visual information, remain suboptimal due to their lack of rigorous reward mechanisms to enforce evidence purity and their inability to perform temporal information supplementation beyond pre-sampled frames. To address this critical gap, we propose a novel evidence-prioritized adaptive framework built upon our core philosophy: "Select Less, Reason More." Our core contribution is the evidence-aware reinforcement learning (EARL) framework, which transforms the model into an active interrogator of evidence. EARL is precisely engineered to dynamically select the most relevant frames and, crucially, to perform localized re-sampling around the selected key frames to access fine-grained temporal detail. Extensive experiments on five demanding video reasoning benchmarks demonstrate that our EARL-trained model achieves new state-of-the-art among open-source Video LLMs, simultaneously learning an effective and high-purity visual evidence selection policy. Impressively, our 7B model achieves 59.8% on LongVideoBench, 69.0% on MVBench and 64.9% on VideoMME. These results highlight the importance of prioritizing evidence purity and the effectiveness of our framework.

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。

↑