InSight-doc: Agentic Visual Perception for Long-Document Understanding

  • 类型:arxiv
  • 标识:2608.10628
  • 链接:https://arxiv.org/abs/2608.10628
  • 主分类:agent
  • 形态:method
  • 被引:0
  • 被引来源:Semantic Scholar
  • S2被引:0
  • 影响力被引:0
  • TLDR:This work proposes InSight-doc, an agentic visual perception framework that treats visual resolution as an adaptive reasoning-time resource that starts from low resolution and selectively zooms into high-resolution regions for finer evidence, without relying on any external retriever.
  • 待LLM分类:否
  • 标题中文:InSight-doc:面向长文档理解的 Agent 视觉感知
  • TLDR中文:本文提出 InSight-doc,一种智能体视觉感知框架,将视觉分辨率视为一种自适应的推理时资源,从低分辨率起步,选择性地放大高分辨率区域以获取更细粒度的证据,且不依赖任何外部检索器。
  • 来源文件
  • /inbox/tom/_candidates/2026-08-12-agent-rag-longcontext-candidates.json
  • /inbox/tom/_candidates/2026-08-13-agent-rag-longcontext-candidates.json
  • [S2 enrich]