Reproduce “Denoising Implicit Variational Inference” 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

Zitong Wang, Zekun Wu, Longlin Yu, Cheng Zhang

DIVI trains a network to predict the marginal score in semi-implicit variational inference from tractable conditional scores, then uses it in the ELBO gradient instead of sampling the reverse conditional. In a 100-dimensional conditioned diffusion task, its coordinate-wise ELM was 75.162, near the SGLD reference’s 75.316, with 335-second time; this diagnostic is not a marginal likelihood or speed guarantee.

Summary by alphaXiv.

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