Reproduce “Same Layout, Different Count: What Number Words Do Past Ten in Text-to-Image Models” 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
Liang Yue, Eric A. Vance
The authors test how reliably five open text-to-image models respond to object counts. They vary the number while holding the object and starting noise fixed, and people count 3,088 images without seeing prompts or models. Small requests generally track the number, but beyond ten behavior differs: some models retain directional influence through 25, while others do not reliably follow larger requests.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.