提升视频理解模型的逻辑一致性和视觉依据，让推理更可信。
Video-R2: Reinforcing Consistent and Grounded Reasoning in Multimodal Language Models
- 用强化学习优化时间对齐与因果连贯性推理
- 在11个基准上同时提升一致性、视觉依赖度和准确率
- 适合需要可靠视频推理的AI研究者与开发者
对动态视觉内容进行推理仍是多模态大模型的核心挑战。现有思维模型虽生成显式推理过程以增强可解释性，但其推理常看似合理却逻辑不一致或弱依赖视觉证据。我们通过两个诊断指标识别并形式化该问题：思考-答案一致性（TAC）衡量推理与答案的一致性，视频注意力分数（VAS）反映推理对视觉与文本线索的依赖程度。在11个视频推理基准上的分析显示，当前模型严重依赖语言先验而非视觉内容。为此，我们提出一种强化学习方法，提升时间精度与推理一致性。该方法结合时间戳感知的监督微调与由新型时间对齐奖励（TAR）指导的组相对策略优化（GRPO）。这一双重后训练阶段促使推理具备时间对齐与因果连贯性。所提出的Video R2模型在多个基准上均实现更高的TAC、VAS与准确率，证明时间对齐与推理连贯性的改进能带来更准确、更可信的视频理解。代码开源：https://github.com/mbzuai-oryx/Video-R2
原文摘要 · Abstract (English)
Reasoning over dynamic visual content remains a central challenge for multimodal large language models. Recent thinking models generate explicit reasoning traces for interpretability; however, their reasoning often appears convincing while being logically inconsistent or weakly grounded in visual evidence. We identify and formalize these issues through two diagnostic metrics: Think Answer Consistency (TAC), which measures the alignment between reasoning and answers, and Video Attention Score (VAS), which captures the extent to which reasoning depends on visual versus textual cues. Analysis across 11 video reasoning benchmarks shows that current models rely heavily on linguistic priors rather than visual content. To address this, we propose a reinforcement learning approach that enhances both temporal precision and reasoning consistency. Our approach combines timestamp aware supervised fine tuning with Group Relative Policy Optimization (GRPO) guided by a novel Temporal Alignment Reward (TAR). This dual step post training stage encourages temporally aligned and causally coherent video reasoning. The resulting model, Video R2, achieves consistently higher TAC, VAS, and accuracy across multiple benchmarks, demonstrating that improvements in temporal alignment and reasoning coherence lead to more accurate and trustworthy video understanding. Code: https://github.com/mbzuai-oryx/Video-R2
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。