Reproduce “Fine-Tuning Generative Models for Extreme Events via CVaR-Penalized Wasserstein Gradient Flows” 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
Thejani Gamage, Hyemin Gu, Zhizhen Zhang, Ziyu Chen, Markos A. Katsoulakis, Luc Rey-Bellet
CVaR-GPA fine-tunes samples from pretrained generative models for heavy-tailed targets. It combines a Lipschitz-regularized KL distribution comparison with a smoothed CVaR correction that emphasizes unusually large radii, without changing model architecture or requiring a target tail-index estimate. Across four targets and seven models, geometric-mean global and tail errors fell 14.0× and 9.8×, respectively; one of 56 comparisons worsened.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.