Reproduce “How VLMs Plan What's Next” 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

Rohit Gandikota, Arnab Sen Sharma, Adam Belfki, David Bau

The authors study how Qwen3-VL-8B plans visual list answers. They find that it binds written answers to image locations and uses the last answer’s position as a spatial cursor to select what to inspect next. In a controlled synthetic task, this cursor predicts the matching letter after panel 4 in 98.4% of cases at layer 20, not general list accuracy.

Summary by alphaXiv.

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