VideoAuto-R1：思考一次、回答两次的视频自动推理
VideoAuto-R1: Video Auto Reasoning via Thinking Once, Answering Twice
提出「按需推理」机制：用初始答案的置信度决定是否触发思维链，可直接迁移到视频问答、视觉定位等 VLM 推理流程中，在几乎不掉点的前提下把平均输出长度压缩 3.3 倍，是降低多模态大模型推理延迟的可复用策略。
原文依据 the model uses the confidence score of the initial answer to determine whether to proceed with 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思维链（CoT）推理已成为多模态大语言模型在视频理解任务上的有力工具。然而，其必要性以及相对于直接回答的优势仍未得到充分探索。本文首先证明，对于经强化学习训练的视频模型，直接回答往往能够持平甚至超越 CoT 的表现，尽管 CoT 会以更高的计算代价给出逐步分析。受此启发，我们提出 VideoAuto-R1，一个采用「按需推理」策略的视频理解框架。在训练阶段，该方法遵循「思考一次、回答两次」的范式：模型先生成一个初始答案，随后进行推理，最后输出一个复核后的答案；两个答案均通过可验证奖励进行监督。在推理阶段，模型利用初始答案的置信度分数来决定是否继续执行推理。在视频问答与视觉定位基准上，VideoAuto-R1 取得了最优准确率，同时效率显著提升，平均响应长度缩短 3.3 倍，例如从 149 个 token 降至仅 44 个 token。此外，我们观察到在感知类任务上思考模式的激活率较低，而在推理密集型任务上激活率较高。这表明显式的语言推理通常有益，但并非总是必要。
英文原文摘要 1362 字符
Chain-of-thought (CoT) reasoning has emerged as a powerful tool for multimodal large language models on video understanding tasks. However, its necessity and advantages over direct answering remain underexplored. In this paper, we first demonstrate that for RL-trained video models, direct answering often matches or even surpasses CoT performance, despite CoT producing step-by-step analyses at a higher computational cost. Motivated by this, we propose VideoAuto-R1, a video understanding framework that adopts a reason-when-necessary strategy. During training, our approach follows a Thinking Once, Answering Twice paradigm: the model first generates an initial answer, then performs reasoning, and finally outputs a reviewed answer. Both answers are supervised via verifiable rewards. During inference, the model uses the confidence score of the initial answer to determine whether to proceed with reasoning. Across video QA and grounding benchmarks, VideoAuto-R1 achieves state-of-the-art accuracy with significantly improved efficiency, reducing the average response length by 3.3x, eg., from 149 to just 44 tokens. Moreover, we observe a low rate of thinking-mode activation on perception-oriented tasks, but a higher rate on reasoning-intensive tasks. This suggests that explicit language-based reasoning is generally beneficial but not always necessary.