[2604.06779] FVD: Inference-Time Alignment of Diffusion Models via Fleming-Viot Resampling

Computer Science > Artificial Intelligence arXiv:2604.06779 (cs) [Submitted on 8 Apr 2026] Title:FVD: Inference-Time Alignment of Diffusion Models via Fleming-Viot Resampling Authors:Shivanshu Shekhar, Sagnik Mukherjee, Jia Yi Zhang, Tong Zhang View a PDF of the paper titled FVD: Inference-Time Alignment of Diffusion Models via Fleming-Viot Resampling, by Shivanshu Shekhar and Sagnik Mukherjee and Jia Yi Zhang and Tong Zhang View PDF HTML (experimental) Abstract:We introduce Fleming-Viot Diffusion (FVD), an inference-time alignment method that resolves the diversity collapse commonly observed in Sequential Monte Carlo (SMC) based diffusion samplers. Existing SMC-based diffusion samplers often rely on multinomial resampling or closely related resampling schemes, which can still reduce diversity and lead to lineage collapse under strong selection pressure. Inspired by Fleming-Viot population dynamics, FVD replaces multinomial resampling with a specialized birth-death mechanism designed for diffusion alignment. To handle cases where rewards are only approximately available and naive rebirth would collapse deterministic trajectories, FVD integrates independent reward-based survival decisions with stochastic rebirth noise. This yields flexible population dynamics that preserve broader trajectory support while effectively exploring reward-tilted distributions, all without requiring value function approximation or costly rollouts. FVD is fully parallelizable and scales efficientl...