VBVR-Pro: A Scalable and Verifiable Suite for Native Visual Reasoning
作者: Junxiang Xu, Ruisi Wang, Fanyi Pu, Maijunxian Wang, Ran Ji, Tongxi Zhou, Chenyang Gu, Jing Zuo, Hongcan Xiao, Yimeng Geng, Wanqi Yin, Wei Chen, Oscar Qian, Zhengan Yan, Ziqi Huang, Haiwen Diao, Liang Pan, Bo Li, Xiangyu Fan, Dezhi Luo, Fengyuan Yu, Zehong Zhao, Qingying Gao, Tinghui Zhu, Yilan Zhang, Jingqi Tong, Pinyuan Feng, Zhengze Jiang, Letian Wang, Ziyu Guo, Renrui Zhang, Jieneng Chen, Sonia Joseph, Constantin Venhoff, Saman Motamed, Mengyue Yang, Chandra Sripada, Alan Yuille, Philip Torr, Lvmin Zhang, Vikash Kumar, Daniel Khashabi, Nikolaus Kriegeskorte, Raphaël Millière, Vincent C. Müller, Anyi Rao, Quan Wang, Ziwei Liu, Dahua Lin, Lei Yang, Hokin Deng, Zhongang Cai
分类: cs.CV, cs.AI, cs.LG, cs.MM, cs.RO
发布日期: 2026-08-26
备注: Homepage: https://video-reason.com/
💡 一句话要点
提出VBVR-Pro以解决可扩展视觉推理训练问题
🎯 匹配领域: 支柱二：RL算法与架构 (RL & Architecture) 支柱八：物理动画 (Physics-based Animation)
关键词: 原生视觉推理 可扩展训练 奖励评分器 多任务强化学习 视觉生成 闭环测试平台 视觉理解 模型迁移
📋 核心要点
- 现有方法在可扩展视觉推理训练任务、可靠反馈和生成基底的比较上存在瓶颈，限制了进展。
- VBVR-Pro通过提供300个程序生成的任务和可验证的奖励评分器，解决了视觉推理的训练和评估问题。
- 在七个外部视觉推理基准上，VBVR-Pro训练的模型表现出强大的迁移能力，且在多任务强化学习中表现优异。
📝 摘要（中文）
原生视觉推理将视觉生成视为推理的媒介，然而在可扩展训练任务、可靠反馈和生成基底的可控比较方面仍存在瓶颈。本文提出VBVR-Pro，一个闭环测试平台，使得通过生成进行的原生视觉推理可训练、可验证、可优化且可实验控制。VBVR-Pro提供300个程序生成的任务，模型在此平台上训练后在七个外部视觉推理基准上表现出强大的迁移能力。此外，VBVR-Pro还提供基于确定性任务特定规则的可验证奖励评分器，显著提升了多任务强化学习的表现。最后，VBVR-Pro支持对30多种图像、视频和交错生成器的控制机制研究，揭示了视觉推理中至关重要的视觉原生轨迹。
🔬 方法详解
问题定义：本文旨在解决原生视觉推理训练中的可扩展性和可靠性问题。现有方法缺乏可控的任务空间和有效的反馈机制，限制了模型的性能提升。
核心思路：VBVR-Pro通过构建一个闭环测试平台，将视觉推理转化为可训练的任务空间，并引入可验证的奖励评分器，以提高训练的有效性和可靠性。
技术框架：VBVR-Pro的整体架构包括任务生成模块、模型训练模块和评估模块。任务生成模块提供300个程序生成的视觉推理任务，模型训练模块使用这些任务进行训练，评估模块则通过可验证的奖励评分器进行性能评估。
关键创新：VBVR-Pro的主要创新在于引入了基于确定性规则的可验证奖励评分器，克服了传统视觉语言模型作为评判者的局限性，提供了更精细的与人类判断的对齐。
关键设计：在设计中，VBVR-Pro采用了多种生成器，包括图像、视频和交错生成器，并通过系统的实验分析，揭示了不同生成方式在视觉推理任务中的表现差异。
🖼️ 关键图片
📊 实验亮点
在实验中，VBVR-Pro训练的模型在七个外部视觉推理基准上表现出强大的迁移能力，显著优于传统方法。此外，使用可验证奖励评分器的多任务强化学习模型在视觉推理任务中表现出更强的后强化学习性能，提升幅度明显。
🎯 应用场景
VBVR-Pro的研究成果在多个领域具有潜在应用价值，包括智能监控、自动驾驶、机器人视觉等。通过提升视觉推理的训练和评估效率，该平台能够推动相关技术的进步，促进更智能的视觉理解系统的开发。
📄 摘要（原文）
Native visual reasoning treats visual generation as the medium of reasoning itself: visual states (i.e. images and videos) are not merely inputs to be understood or outputs to be rendered, but first-class substrates for problem solving beyond language. Yet progress remains bottlenecked by the lack of scalable training tasks, reliable feedback, and controlled comparisons across generative substrates. In this work, we introduce VBVR-Pro, a closed-loop testbed that makes native visual reasoning through generation trainable, verifiable, optimizable, and experimentally controllable. 1) Task scaling. VBVR-Pro turns visual reasoning into a controlled task space of 300 procedurally generated tasks. Models trained on VBVR-Pro show strong transfer beyond the proposed suite across seven external visual reasoning benchmarks such as RISE-Video, MME-CoF-Pro, and BabyVision. 2) Verifiable rewards. VBVR-Pro provides verifiable reward scorers for task-grounded evaluation. Through a systematic study of leading MLLMs as judges, we identify recurring failure modes of the prevalent VLM-as-a-judge paradigm. In contrast, the proposed scorers are grounded in deterministic, task-specific rules, achieve fine-grained alignment with human judgments. Importantly, they serve as reliable reward signals for large-scale multi-task reinforcement learning and demonstrate stronger post-RL performance across visual reasoning tasks. 3) Mechanism study. VBVR-Pro enables controlled modality studies across more than 30 image, video, and interleaved generators. Our analysis shows that video generation remains strongest for tasks requiring persistent spatiotemporal state tracking, while interleaved generation provides a compute-efficient alternative. Critically, ablations and probing suggest the presence of vision-native trajectories that are crucial to visual reasoning. We release all data, models, scorers, and code.