Reproduce “On the Limits of Latent Reuse in Diffusion 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
Yifeng Yu, Lu Yu
The authors analyze when a frozen source representation prevents a diffusion model from recovering target scores. They model source and target signals as noisy low-dimensional subspaces and prove a lower bound with contributions from missed target signal and ambient-noise score error. Containment removes the signal-mismatch contribution, but not necessarily ambient-noise error; the result assumes noisy linear subspaces.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.