arXiv cs.AI· 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·· 3 小时前AI 评分46
Zooming without Zooming v2：从看到细节到连接证据的泛化多模态感知
Zooming without Zooming v2: From Seeing Details to Connecting Evidence for Generalizable Multimodal Perception
AI 导读
研究团队提出数据合成框架 EvidenceConnect，利用单张图像内相关区域的放大视图生成需比较、聚合或串联视觉证据的问题，并通过多 VLM 一致性校验与区域消融测试保证数据质量。基于该数据用 RLVR 训练得到 ZwZ2 模型，虽数据未针对下游 benchmark 定制，却在细粒度与空间感知上提升，并在具身空间推理、多模态长上下文理解、智能体任务及通用视觉推理上获得进一步增益。
正文 · 原文
Authors: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
Abstract: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.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG) |
| Cite as: | arXiv:2610.37037 [cs.CV] |
| (or arXiv:2610.37037v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.37037 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Lai Wei [view email]
[v1]
Tue, 27 Oct 2026 11:11:53 UTC (3,416 KB)
来源：arXiv cs.AI · arxiv.org