Reproduce “Content-Preserving Principal Feature Blending for Training-Free Style Personalization” 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

Le-Hoang Bui, Hoang-Nguyen Cao, Trung-Nghia Le, Minh-Triet Tran, Dinh-Khoi Vo

The authors introduce CoPFB for reference-based style personalization in visual autoregressive generation. It estimates content directions from a clean prompt-guided generation, then filters aligned directions from the reference-style edit before injection. On 1,900 Parti–FineStyle pairs with Infinity-2B, its hybrid schedule raises text similarity and harmonic score versus PFB+SAC, while lowering image similarity. This improves measured balance, not style fidelity.

Summary by alphaXiv.

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