{"ID":2866249,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_url":"https://arxiv.org/abs/2509.21655","arxiv_id":"2509.21655","title":"DriftLite: Lightweight Drift Control for Inference-Time Scaling of Diffusion Models","abstract":"We study inference-time scaling for diffusion models, where the goal is to adapt a pre-trained model to new target distributions without retraining. Existing guidance-based methods are simple but introduce bias, while particle-based corrections suffer from weight degeneracy and high computational cost. We introduce DriftLite, a lightweight, training-free particle-based approach that steers the inference dynamics on the fly with provably optimal stability control. DriftLite exploits a previously unexplored degree of freedom in the Fokker-Planck equation between the drift and particle potential, and yields two practical instantiations: Variance- and Energy-Controlling Guidance (VCG/ECG) for approximating the optimal drift with minimal overhead. Across Gaussian mixture models, particle systems, and large-scale protein-ligand co-folding problems, DriftLite consistently reduces variance and improves sample quality over pure guidance and sequential Monte Carlo baselines. These results highlight a principled, efficient route toward scalable inference-time adaptation of diffusion models. Our source code is publicly available at https://github.com/yinuoren/DriftLite.","short_abstract":"We study inference-time scaling for diffusion models, where the goal is to adapt a pre-trained model to new target distributions without retraining. Existing guidance-based methods are simple but introduce bias, while particle-based corrections suffer from weight degeneracy and high computational cost. We introduce Dri...","url_abs":"https://arxiv.org/abs/2509.21655","url_pdf":"https://arxiv.org/pdf/2509.21655v2","authors":"[\"Yinuo Ren\",\"Wenhao Gao\",\"Lexing Ying\",\"Grant M. Rotskoff\",\"Jiequn Han\"]","published":"2025-09-25T22:21:59Z","proceeding":"cs.LG","tasks":"[\"cs.LG\",\"stat.ML\"]","methods":"[\"Diffusion Model\",\"Generative Adversarial Network\"]","has_code":false,"code_links":[{"ID":609355,"CreatedAt":"2026-06-01T04:54:23.091178241Z","UpdatedAt":"2026-06-01T04:54:23.091178241Z","DeletedAt":null,"paper_id":2866249,"paper_url":"https://arxiv.org/abs/2509.21655","paper_title":"DriftLite: Lightweight Drift Control for Inference-Time Scaling of Diffusion Models","repo_url":"https://github.com/yinuoren/DriftLite","is_official":false,"mentioned_in_paper":false,"mentioned_in_github":true,"github_stars":0}]}