SFBD Flow: A Continuous-Optimization Framework for Training Diffusion Models with Noisy Samples
This addresses privacy issues in diffusion models for AI practitioners, but it is incremental as it builds on prior SFBD work.
The paper tackles the problem of training diffusion models with privacy concerns by introducing SFBD flow, a continuous-optimization framework that eliminates manual coordination in iterative denoising, and demonstrates that Online SFBD outperforms baselines across benchmarks.
Diffusion models achieve strong generative performance but often rely on large datasets that may include sensitive content. This challenge is compounded by the models' tendency to memorize training data, raising privacy concerns. SFBD (Lu et al., 2025) addresses this by training on corrupted data and using limited clean samples to capture local structure and improve convergence. However, its iterative denoising and fine-tuning loop requires manual coordination, making it burdensome to implement. We reinterpret SFBD as an alternating projection algorithm and introduce a continuous variant, SFBD flow, that removes the need for alternating steps. We further show its connection to consistency constraint-based methods, and demonstrate that its practical instantiation, Online SFBD, consistently outperforms strong baselines across benchmarks.