Reproduce “MLLM-Guided Autonomous Object Detection for Open-World Understanding” 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

Furkan Mumcu, Michael J. Jones, Anoop Cherian, Yasin Yilmaz

LAOD lets an MLLM generate image-specific object names, then passes them with the image to an open-vocabulary detector for localization. On COCO-Val, using Qwen2.5-VL-7B to generate vocabulary for YOLO-World raises CAAP from 0.1422 to 0.3004 and SemLoc-AP from 0.0359 to 0.1875 versus direct detection. This compares full pipelines, not the effect of modularity alone; prompts still guide scope.

Summary by alphaXiv.

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