Reproduce “Worth Another Look:Constrained World-Model Planning for Visual Inspection” 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
Ziyue Kang, Zixi Qi, Yuchen Dai, Nana Lv, Xirong Cheng, Liang Liu, Ying Li, Haoyan Wei and 1 more
The authors present a planner that lets frozen detectors inspect images adaptively under sensing and compute limits. It uses a recurrent world model and constrained lookahead to choose class queries, crops, image enhancements, or stopping, while symbolic rules enforce feasibility. On RUOD, at budget B = 2, three-step planning reaches 0.404 AP50:95, versus 0.345 at one step; gains vary across datasets.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.