Reproduce “CRNDiff: Count-Native Diffusion Framework via Chemical Reaction Networks” 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
Yuxuan Qiu, Praful Gagrani, Tetsuya J Kobayashi
CRNDiff generates integer gene-expression profiles for cell types with birth–death dynamics and two-stage steering. It first tilts reverse sampling using target-to-generator one-gene marginal ratios, then uses particle weighting and resampling to correct joint differences while keeping the generator frozen. For rare neuronal cells, purity is 0.971 ± 0.012 versus MDLM’s 0.784 ± 0.063; purity measures labeling, not distributional equality.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.