VLMs 是视频推理的好老师：通过自适应测试时优化
VLMs are Good Teachers for Video Reasoning via Adaptive Test-Time Optimization
AI总结：
提出将视觉语言模型（VLM）作为“教师”，通过提取任务规则并设计可微分奖励，指导视频生成模型（VGM）在测试时在线优化轻量级 LoRA 模块，从而提升视频推理的泛化能力。
摘要：
最近的“视频推理”范式利用视频生成模型（VGM）生成时间连贯的视觉轨迹来完成推理任务。尽管最先进的 VGM 在视觉质量上表现出色，但它们往往难以理解和遵循任务特定规则，导致在各种推理场景中出现逻辑失败。现有工作尝试利用视觉语言模型（VLM）作为问题预求解器，为 VGM 生成或细化文本指导。然而，文本描述无法捕捉复杂的时空细节，并且 VGM 即使有有效计划，也常常难以忠实执行细粒度或长尾指令。尽管 VLM 作为求解器存在困难，但它们具备强大的感知能力，可以评估过程约束满足和最终目标达成。利用这一优势，我们引入了一种范式转变，将 VLM 的角色转变为“教师”。具体来说，VLM 教师提取任务特定规则以制定可微分奖励，通过测试时在线优化轻量级 LoRA 模块来指导 VGM 推理器。该策略实现了自适应测试时优化，并将推理能力扩展到 VGM 的内在边界之外。在符号（VBVR-Bench）和通用（RULER-Bench）视频推理基准上的评估表明，所提出的方法平均性能提升 16.7 个百分点，在可比测试时成本下，大幅优于 VLM-as-Solver 范式（+0.4 个百分点）和 Best-of-N 缩放（+2.2 个百分点）。这些发现表明，将 VLM 作为测试时教师集成，为实现可泛化的视频推理提供了一种有前景的范式。项目页面：https://VLM-as-Teacher.github.io/
英文摘要：
The recent "Reasoning with Video" paradigm utilizes Video Generation Models (VGMs) to generate temporally coherent visual trajectories to complete reasoning tasks. Although state-of-the-art VGMs excel at visual quality, they often struggle to understand and follow task-specific rules, leading to logical failures across diverse reasoning scenarios. Existing efforts try to utilize Vision-Language Models (VLMs) as problem pre-solvers to produce or refine textual guidance for the VGM. However, textual descriptions fail to capture intricate spatiotemporal details, and VGMs often struggle to faithfully execute fine-grained or long-tail instructions even with a valid plan. While VLMs struggle as solvers, they possess strong perception capabilities to evaluate process-constraint satisfaction and final-goal achievement. Leveraging this strength, we introduce a paradigm shift that transitions the role of VLMs to "teachers". Specifically, a VLM teacher extracts task-specific rules to formulate differentiable rewards, guiding a VGM Reasoner via test-time online optimization of a lightweight LoRA module. This strategy enables adaptive test-time optimization and extends the reasoning capabilities beyond the VGM's intrinsic boundaries. Evaluations on symbolic (VBVR-Bench) and general-purpose (RULER-Bench) video reasoning benchmarks show that the proposed method yields a 16.7-point average performance gain, outperforming the VLM-as-Solver paradigm (+0.4 points) and Best-of-N scaling (+2.2 points) by a large margin at comparable test-time cost. These findings reveal that integrating VLMs as test-time teachers offers a promising paradigm for achieving generalizable video reasoning. Project Page: https://VLM-as-Teacher.github.io/
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。
