Video-CoE：通过事件链强化视频事件预测
Video-CoE: Reinforcing Video Event Prediction via Chain of Events
论文提出的 Chain of Events 范式通过构建时序事件链来隐式引导 MLLM 关注视觉内容与事件间的逻辑关联，这种将推理链结构嵌入视频时序建模的思路，可迁移到视频问答、动作预测等需要细粒度时序推理的视觉任务中。
原文依据 VEP requires the model to perform fine-grained temporal modeling of videos and establish logical relationships between videos and future events
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论尽管多模态大语言模型（MLLM）在各类视频任务中的应用取得了进展，视频事件预测（VEP）仍相对未被充分探索。VEP 要求模型对视频进行细粒度时序建模，并建立视频与未来事件之间的逻辑关系，而当前 MLLM 仍难以做到。在本工作中，我们首先对当前领先的 MLLM 在 VEP 任务上进行了全面评估，揭示了其预测不准确的原因，包括缺乏对未来事件预测的逻辑推理能力以及对视觉信息利用不足。为应对这些挑战，我们提出了事件链（Chain of Events, CoE）范式，该范式构建时序事件链，以隐式地促使 MLLM 关注视觉内容以及视频与未来事件之间的逻辑联系，并通过多种训练协议激励模型的推理能力。在公开基准上的实验结果表明，我们的方法优于领先的开源和商业 MLLM，在 VEP 任务上建立了新的最优性能。
英文原文摘要 1108 字符
Despite advances in the application of MLLMs for various video tasks, video event prediction (VEP) remains relatively underexplored. VEP requires the model to perform fine-grained temporal modeling of videos and establish logical relationships between videos and future events, which current MLLMs still struggle with. In this work, we first present a comprehensive evaluation of current leading MLLMs on the VEP task, revealing the reasons behind their inaccurate predictions, including lack of logical reasoning ability for future events prediction and insufficient utilization of visual information. To address these challenges, we propose **C**hain **o**f **E**vents (**CoE**) paradigm, which constructs temporal event chains to implicitly enforce MLLM focusing on the visual content and the logical connections between videos and future events, incentivizing model's reasoning capability with multiple training protocols. Experimental results on public benchmarks demonstrate that our method outperforms both leading open-source and commercial MLLMs, establishing a new state-of-the-art on the VEP task.