Reproduce “ScoreDial: Precise Control of Visual Attributes in Image Generation via Regressor Guidance from Limited Human Ratings” 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
Jianbo Tang, Keita Ishii, Hiroki Sakamoto, Manabu Kikuchi, Katsuma Kitamura, Chiho Semba, Kenta Nishiyama, Teppei Mori and 2 more
ScoreDial steers a pretrained image generator toward requested levels of human-rated attributes, using ratings from a few hundred images to train a score model. It combines requested-value conditioning with feedback that corrects score errors during sampling. Across four attributes, an independent auditor shows greater movement across requested values than competing methods; human evidence covers only tire and soap, and score agreement alone does not establish perceived change.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.