Reproduce “Probabilistic Precipitation Nowcasting with Rectified Flow Transformers” 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
Johannes Schusterbauer, Jannik Wiese, Nick Stracke, Timy Phan, Björn Ommer
FREUD enables probabilistic rainfall nowcasting by pairing a frame-independent encoder with a generative decoder that reconstructs video jointly, allowing one latent representation to produce multiple reconstructions. In SEVIR tests, its T-reg. variant’s reconstruction variance correlates with precipitation intensity (r = 0.997); LSM-L achieves CRPS 0.0190 and SSIM 0.7841, but localization scores remain below CasCast’s.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.