VICO: Visual Environments Co-Evolving for Vision-Language Model Reasoning

Computer Science > Computer Vision and Pattern Recognition arXiv:2610.10782 (cs) [Submitted on 7 Oct 2026] Title: VICO: Visual Environments Co-Evolving for Vision-Language Model Reasoning Authors: Meng Lu, Ligeng Zhu, Olivia Xiao, Yuchen Zhuang, Zihan Wang, Kuncheng Wu, Bangya Liu, Yu Wang, Charles Fleming, Wenqi Shi, Xuan Wang.

计算机科学 > 计算机视觉与模式识别 arXiv:2610.10782 (cs) [提交于 2026 年 10 月 7 日] 标题：VICO：用于视觉语言模型推理的视觉环境协同进化 作者：Meng Lu, Ligeng Zhu, Olivia Xiao, Yuchen Zhuang, Zihan Wang, Kuncheng Wu, Bangya Liu, Yu Wang, Charles Fleming, Wenqi Shi, Xuan Wang。

Abstract: Reinforcement learning with verifiable rewards (RLVR) has become a standard recipe for post-training vision-language models (VLMs), but it typically assumes a static training environment. As the actor improves, fixed tasks drift out of its learning frontier: many become trivial, others remain unsolvable; and the learning signal collapses.

摘要：带有可验证奖励的强化学习（RLVR）已成为视觉语言模型（VLM）训练后阶段的标准方法，但它通常假设训练环境是静态的。随着智能体（actor）的提升，固定的任务会偏离其学习前沿：许多任务变得微不足道，而另一些则仍然无法解决；导致学习信号崩溃。

We argue that VLM post-training should evolve the visual environment alongside the actor, not just the actor itself. We propose VICO, a co-evolutionary framework in which an actor and an Environment-as-Rewriter (EnvRewriter) are trained jointly: the EnvRewriter edits verifiable image-side structures, such as scene graphs, chart tables, or protected region masks, and re-renders them to produce label-valid training samples whose difficulty is calibrated to the actor’s current ability through a pass-rate-based reward.

我们认为，VLM 的训练后阶段应该让视觉环境与智能体共同进化，而不仅仅是智能体本身。我们提出了 VICO，这是一个协同进化框架，其中智能体和“环境即重写器”（EnvRewriter）被联合训练：EnvRewriter 编辑可验证的图像侧结构（如场景图、图表表格或受保护区域掩码），并重新渲染它们以生成标签有效的训练样本，其难度通过基于通过率的奖励来校准，以匹配智能体当前的能力。

This loop continuously realigns task difficulty with actor capability without any additional human annotation. Across nine multimodal benchmarks spanning mathematical reasoning and visually grounded understanding, VICO-8B improves over its base model by up to +5.0% on out-of-domain tasks, surpasses the strongest self-evolution and text-editing co-evolution baselines by +4.3% and +8.4% respectively, and stays comparable to chart-specialized RLVR methods using 16-160 times fewer labeled samples.

这一循环在无需任何额外人工标注的情况下，持续将任务难度与智能体能力重新对齐。在涵盖数学推理和视觉基础理解的九个多模态基准测试中，VICO-8B 在域外任务上比其基础模型提升了高达 +5.0%，分别超过了最强的自我进化和文本编辑协同进化基线 +4.3% 和 +8.4%，并且在使用少 16 到 160 倍标注样本的情况下，保持了与图表专用 RLVR 方法相当的性能。

By shifting from human-labeled supervision to image-editing co-evolution, VICO offers a scalable path beyond static-corpus RLVR for visual reasoning.

通过从人工标注监督转向图像编辑协同进化，VICO 为视觉推理提供了一条超越静态语料库 RLVR 的可扩展路径。