Reproduce “Agentic Ensemble with Vision-Language Models for Visual Anomaly Detection” 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

Shangquan Sun, Wenjun Luo, Xu He, Yoonsoo Lee, Shijian Lu

AnomalyAgent uses a vision-language model to select predefined ways to fuse fixed detectors’ image scores and pixel maps when evidence conflicts, rather than generating scores or masks. On MVTec AD’s CLIP-based pool, image AUROC rises from 98.80% to 99.25% and pixel AP from 64.43% to 72.31%; this exceeds equal-weight fusion, while matching AdaCLIP alone in pixel AP.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.