AI RESEARCH
Lookahead Sample Reward Guidance for Test-Time Scaling of Diffusion Models
arXiv CS.AI
•
ArXi:2602.03211v2 Announce Type: replace-cross Diffusion models have nstrated strong generative performance; however, generated samples often fail to fully align with human intent. This paper studies an efficient test-time scaling method for sampling from regions with higher human-aligned reward values. Existing methods for computing the expected future reward (EFR) face important limitations: backward rollout incurs prohibitively high sampling costs, while Tweedie-based approaches, including Sequential Monte Carlo and gradient guidance, suffer from bias and inherent sampling issues.