Lune

CVPR Findings2026

Video Reasoning Without Training

Deepak Sridhar, Kartikeya Bhardwaj, Jeya Pradha Jeyaraj, Nuno Vasconcelos, Ankita Nayak, Harris Teague

问 Lune

摘要

Video reasoning using Large Multimodal Models (LMMs) relies on costly reinforcement learning (RL) and verbose chainof-thought, resulting in substantial computational overhead during both training and inference. Moreover, the mechanisms that control the thinking process in these reasoning models are very limited. In this paper, we use the entropy of the model's output distribution as a signal to study and guide reasoning behavior. We discover that high-quality models exhibit a characteristic pattern of micro-exploration and micro-exploitation cycles, followed by a later entropy peak (i.e., longer thinking) and a lower final entropy, indicating more deliberate exploration and confident convergence (i.e., avoid excessive randomness while the model is exploring or thinking through an answer). We then use these novel, theoretically-grounded insights to introduce V-Reason (Video-Reason), an inference-time optimization method that adapts the value cache of the LMM through a lightweight, trainable controller. Our proposed controller is guided by an entropy-based objective, to tune the model's behavior directly at inference, without using any RL or supervised fine-tuning. Our experiments show that V-Reason significantly outperforms the base instruction-tuned models on many video reasoning datasets, narrowing the gap with RL models to within 0.6% accuracy on average. We achieve this without any training, while offering efficiency benefits: V-Reason uses 58.6% fewer tokens than the RL model.

问 Lune

几分钟读透这篇 Paper,不用再花一下午。

Lune 能读完全文,也能读引用它的顶会 Paper,讲清它的结论、证据和局限,每一点都注明出自哪篇 Paper。不用硬啃,也不靠猜。

可以从这些问题问起

Lune 可以查阅

全文 · 相关的 Paper

获取答案

免费开始,无需绑卡

脉络

  • Large Language Models are Zero-Shot Reasoners 引用了 Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
  • Large Language Models as Optimizers 引用了 Large Language Models are Zero-Shot Reasoners
  • Large Language Models as Optimizers 引用了 Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
  • Video-R1: Reinforcing Video Reasoning in MLLMs 引用了 Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
它引用的引用它的从左到右按年份排列,引用越多,位置越高。

它引用的 Paper29

  1. Chain-of-Thought Prompting Elicits Reasoning in Large Language ModelsJason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma 等NeurIPS 2022 · 22,562 次引用
  2. Large Language Models are Zero-Shot ReasonersTakeshi Kojima, Shixiang Shane Gu, Machel Reid, Yutaka Matsuo 等NeurIPS 2022 · 8,168 次引用
  3. Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?Zhiqi Chen, Rui Lu, Andrew Zhao, Zhaokai Wang 等NeurIPS 2025 · 1,109 次引用
  4. Large Language Models as OptimizersChengrun Yang, Xuezhi Wang, Yifeng Lu, Hanxiao Liu 等ICLR 2024 · 817 次引用
  5. Video-R1: Reinforcing Video Reasoning in MLLMsKaituo Feng, Kaixiong Gong, Bohao Li, Zonghao Guo 等NeurIPS 2025 · 528 次引用

相关 Paper

  1. ARES: Multimodal Adaptive Reasoning via Difficulty-Aware Token-Level Entropy ShapingShuang Chen, Hangyu Guo, Yimeng Ye, Shijue Huang 等ICLR 2026 · 23 次引用
  2. VideoAuto-R1: Video Auto Reasoning via Thinking Once, Answering TwiceShuming Liu, Mingchen Zhuge, Changsheng Zhao, Jun Chen 等CVPR 2026 · 18 次引用
  3. VAST: Video Ability-Stratified Taxonomy for Data-Efficient Video ReasoningZhongan Wang, Xiaoyu Wen, Lingxiao Du, Kun Li 等CVPR 2026
  4. Temporal-Aware Reasoning Optimization for Video Temporal GroundingMinghang Zheng, Zihao Yin, YI YANG, Yuxin Peng 等ICML 2026
  5. When Thinking Drifts: Evidential Grounding for Robust Video ReasoningRomy Luo, Zihui Xue, Alex Dimakis, Kristen GraumanNeurIPS 2025 · 21 次引用
黄昏的海面,两侧是细线勾勒的悬崖