发表机构
BUAA; CUHK; NTU; THU; HFUT（北京航空航天大学; 香港中文大学; 南洋理工大学; 清华大学; 合肥工业大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究提出以视觉定位证据图（VGEG）为核心的统一多模态深度研究智能体OneSearch-VL，构建相关数据集与基准，在图像、视频深度研究任务上较Qwen3-VL-8B取得显著性能提升。
AI 中文摘要
单图像、多图像及视频的深度研究需要不同的视觉操作，但共享视觉定位、外部检索与事实合成的工作流程。核心挑战在于保留局部视觉锚点、实体关系、源支持事实及答案生成操作之间的依赖关系。我们提出OneSearch-VL，这是一个以视觉定位证据图（Visually Grounded Evidence Graph, VGEG）为核心的统一智能体，将这些依赖关系编码为数据构建、过程监督及操作级评估的共享任务级参考。基于VGEG的数据引擎构建并验证多图像与视频问题，过滤专家轨迹。利用这些数据，我们分别组装用于监督微调（SFT）的OneSearch-VL-SFT-110K和用于强化学习（RL）的OneSearch-VL-RL-10K。我们进一步从VGEG标注中推导证据感知视觉定位规则奖励（Evidence-aware Visual-Grounded Rubric reward, EVGR），以在RL过程中监督证据可追溯性与视觉定位。为进行细粒度评估，我们构建OneSearch-MI-Bench与OneSearch-Video-Bench，按其VGEG编码的研究操作组织问题。实验表明，OneSearch-VL-8B在两个新基准上较带工具访问的Qwen3-VL-8B分别提升20.2和17.6个百分点，同时在7个图像基准及VideoDR上也取得显著增益。项目代码仓库：this https URL
英文摘要
Single-image, multi-image, and video deep research require different visual operations but share a workflow of visual grounding, external retrieval, and fact composition. A key challenge is to preserve the dependencies linking localized visual anchors, entity relations, source-supported facts, and answer-producing operations. We introduce OneSearch-VL, a unified agent centered on the Visually Grounded Evidence Graph (VGEG), which encodes these dependencies as a shared task-level reference for data construction, process supervision, and operation-level evaluation. Our VGEG-based data engine constructs and verifies multi-image and video questions and filters expert trajectories. Using these data, we assemble OneSearch-VL-SFT-110K and OneSearch-VL-RL-10K for SFT and RL, respectively. We further derive the Evidence-aware Visual-Grounded Rubric reward (EVGR) from VGEG annotations to supervise evidence traceability and visual grounding during RL. For fine-grained evaluation, we construct OneSearch-MI-Bench and OneSearch-Video-Bench, organizing questions by the research operations encoded in their VGEGs. Experiments show that OneSearch-VL-8B improves over Qwen3-VL-8B with tool access by 20.2 and 17.6 percentage points on the two new benchmarks, respectively, while also achieving substantial gains across 7 image benchmarks and VideoDR. Project repository: https://github.com/appletea233/OneSearch-VL
