← back to paper
arxiv: 2604.06779 · 2 revisions
VASR: Variance-Aware Systematic Resampling for Reward-Guided Diffusion