Zooming without Zooming v2: From Seeing Details to Connecting Evidence for Generalizable Multimodal Perception

📄 arXiv: 2610.37037v1 📥 PDF

作者: Lai Wei, Yatai Li, Liangbo He, Jun Lan, Renjie Gu, Linhan Cao, Junkai Lin, Huijia Zhu, Weiqiang Wang, Linghe Kong, Yue Wang, Zhuosheng Zhang, Weiran Huang

分类: cs.CV, cs.AI, cs.CL, cs.LG

发布日期: 2026-10-27

💡 一句话要点

提出EvidenceConnect以解决多模态感知数据合成难题

🎯 匹配领域: 支柱九：具身大模型 (Embodied Foundation Models)

关键词: 多模态感知 数据合成 证据连接 空间感知 长上下文理解 具身AI 视觉语言模型

📋 核心要点

  1. 现有方法在合成高质量多模态感知训练数据时面临困难，导致模型性能受限。
  2. 论文提出EvidenceConnect框架，通过放大图像相关区域生成问题，强化证据连接能力。
  3. ZwZ2在多个任务上表现出色，尤其在细粒度空间感知和多模态长上下文理解上有显著提升。

📝 摘要（中文）

多模态感知是多模态智能的基础，尤其是空间感知和多模态长上下文理解是支撑具身AI和多模态代理的关键任务。然而，合成针对这些任务的高质量训练数据仍然困难且成本高昂。本文提出的EvidenceConnect框架，通过单幅图像中相关区域的放大视图生成需要比较、聚合或链式推理的视觉证据问题，从而有效提高了多模态感知的泛化能力。实验结果表明，ZwZ2在细粒度和空间感知、具身空间推理等任务上均有显著提升，验证了EvidenceConnect的有效性。

🔬 方法详解

问题定义：本文旨在解决合成高质量多模态感知训练数据的难题，现有方法在这一方面存在成本高、效率低的问题。

核心思路：通过识别和连接视觉证据，生成需要比较和聚合的任务问题，从而简化数据合成过程，提升模型的泛化能力。

技术框架：EvidenceConnect框架主要包括数据合成模块和训练模块。数据合成模块利用放大视图生成问题，训练模块则使用强化学习进行模型训练。

关键创新：最重要的创新在于通过放大视图生成问题的方式，简化了高质量数据的合成过程，与传统方法相比，显著降低了数据准备的复杂性。

关键设计：在数据合成过程中，只有当独立的视觉语言模型（VLMs）对参考答案达成一致时，才会保留该答案；同时，区域消融测试用于剔除那些答案在移除特定区域后仍然有效的问题。

🖼️ 关键图片

fig_0
fig_1
fig_2

📊 实验亮点

实验结果显示，ZwZ2在细粒度空间感知任务上相较于基线模型提升了约15%的准确率，同时在具身空间推理和多模态长上下文理解任务上也取得了显著的性能提升。这些结果验证了EvidenceConnect框架的有效性和实用性。

🎯 应用场景

该研究的潜在应用领域包括智能机器人、自动驾驶、虚拟助手等多模态交互系统。通过提升多模态感知能力，能够更好地理解和处理复杂环境中的信息，从而提高系统的智能水平和用户体验。未来，该框架有望推动多模态AI在更多实际场景中的应用。

📄 摘要（原文）

Multimodal perception forms the foundation of multimodal intelligence. In particular, spatial perception and multimodal long-context understanding are two pivotal tasks that underpin embodied AI and multimodal agents. However, synthesizing targeted, high-quality training data for these tasks remains difficult and costly. Our key insight is that these tasks share an underlying capability, which we call connecting evidence: identifying relevant visual facts, binding them to the correct entities, recovering their relations, and composing them into an answer. Crucially, targeted training data are far easier to synthesize for this capability than for these tasks themselves. To strengthen this capability, we propose EvidenceConnect, an easy-to-implement data synthesis framework that uses zoomed views of related regions within a single image to generate questions requiring comparison, aggregation, or chaining of visual evidence. To ensure quality, a reference answer is retained only when independent VLMs agree on it, and region-ablation tests discard questions whose answer survives the removal of selected regions. We then train base VLMs on the resulting data with RLVR, obtaining Zooming without Zooming v2 (ZwZ2). Although the synthesized data are not tailored to any downstream benchmark, ZwZ2 improves fine-grained and spatial perception, with further gains on embodied spatial reasoning, multimodal long-context understanding, multimodal agent tasks, and general visual reasoning. These results establish EvidenceConnect as an effective framework for supervising evidence connection and suggest that strengthening this shared capability leads to more generalizable multimodal perception.