Reproduce “Towards High-Resolution Visual Perception via Hierarchical Entity Exploration” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Ziyu Ma, Shidong Yang, Yuxiang Ji, Yiming Hu, Tongwen Huang, Yong Wang, Jianfei Cai, Xiangxiang Chu
HEE searches high-resolution images by exploring detected object-based regions rather than fixed crops. It scores candidates using question-region similarity and model confidence, expands promising regions, and backtracks when needed. On Visual Probe Hard, Qwen2.5-VL-7B achieves 50.0% accuracy, compared with 41.5% for ZoomEye and 34.0% for RAP. This result supports HEE in that setting, but does not guarantee correct answers.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.