ReMoT：基于运动对比三元组的强化学习
ReMoT: Reinforcement Learning with Motion Contrast Triplets
用规则化流程从视频元标注自动构造运动对比三元组，并以 GRPO 替代 SFT 训练 VLM 判别细微运动差异（如相反方向），为视频时序理解与具身/驾驶场景的时空一致性提供可复用的数据构造与强化学习后训练方案。
原文依据 Group Relative Policy Optimization, which we empirically validate, yields optimal performance and data efficiency for learning this contrastive reasoning
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论本文提出 ReMoT，一种统一的训练范式，用于系统性地解决视觉语言模型在时空一致性方面的根本缺陷——这是导航、机器人与自动驾驶中的关键失效点。ReMoT 集成了两个核心组件：(i) 一个基于规则的自动化框架，用于生成 ReMoT-16K，这是一个由视频元标注派生的大规模（16.5K 个三元组）运动对比数据集，优于昂贵的人工生成或基于模型的生成方式。(ii) 分组相对策略优化，我们通过实验验证，它在学习这种对比推理时能达到最优的性能与数据效率，远超标准的监督微调。我们还构建了首个面向细粒度运动对比三元组的基准，用于衡量视觉语言模型对细微运动属性（例如相反方向）的判别能力。所得模型在我们的新基准以及多个标准视觉语言模型基准上取得了 SOTA 性能，在时空推理任务上实现了高达 25.1 的性能跃升。
英文原文摘要 1028 字符
We present ReMoT, a unified training paradigm to systematically address the fundamental shortcomings of VLMs in spatio-temporal consistency--a critical failure point in navigation, robotics, and autonomous driving. ReMoT integrates two core components: (i) A rule-based automatic framework that generates ReMoT-16K, a large-scale (16.5K triplets) motion-contrast dataset derived from video meta-annotations, surpassing costly manual or model-based generation. (ii) Group Relative Policy Optimization, which we empirically validate, yields optimal performance and data efficiency for learning this contrastive reasoning, far exceeding standard Supervised Fine-Tuning. We also construct the first benchmark for fine-grained motion contrast triplets to measure a VLM's discrimination of subtle motion attributes (e.g., opposing directions). The resulting model achieves SOTA performance on our new benchmark and multiple standard VLM benchmarks, culminating in a remarkable 25.1 performance leap on spatio-temporal reasoning tasks.