AI RESEARCH

Inference-Time Scaling of Discrete Diffusion Models via Importance Weighting and Optimal Proposal Design

arXiv CS.LG

ArXi:2505.22524v4 Announce Type: replace Discrete diffusion models have become highly effective across various domains. However, real-world applications often require the generative process to adhere to certain constraints. To this end, we propose a Sequential Monte Carlo (SMC) framework that enables scalable inference-time control of discrete diffusion models through principled importance weighting and optimal proposal construction.