PixelEyes: Decoupling Perception and Reasoning for Pinpoint Visual Evidence Seeking

  • 类型:arxiv
  • 标识:2607.00115
  • 链接:https://arxiv.org/abs/2607.00115
  • 主分类:agent
  • 形态:method
  • 被引:1
  • 被引来源:Semantic Scholar
  • S2被引:1
  • OpenAlex被引:0
  • 影响力被引:0
  • TLDR:This paper explores multi-turn visual reasoning and observes that MLLMs repeatedly fail to localize the target, leading to long, redundant trajectories, and proposes PixelEyes, a multi-turn visual reasoning agent that explicitly decouples reasoning from perception.
  • OpenAlex ID:W7167094583
  • OpenAlex DOI:10.48550/arxiv.2607.00115
  • DOI:10.48550/arxiv.2607.00115
  • DOI来源:OpenAlex
  • 开放获取:green
  • 开放获取链接:https://doi.org/10.48550/arxiv.2607.00115
  • OpenAlex更新:2026-07-19
  • 待LLM分类:否
  • 标题中文:PixelEyes:解耦感知与推理以实现精准视觉证据定位
  • TLDR中文:本文探索多轮视觉推理,观察到 MLLM 反复无法定位目标,导致冗长的推理轨迹,进而提出 PixelEyes,一种将推理与感知显式解耦的多轮视觉推理 Agent。
  • 来源文件
  • /inbox/tom/_candidates/2026-07-02-agent-rag-longcontext-candidates.json
  • [S2 enrich]
  • [OpenAlex backfill]