Integration Matters: Rollout-Based Training for Constrained Diffusion Models
For practitioners using diffusion models under feasibility constraints, this method reduces distribution shift and tuning costs, offering a more principled fine-tuning approach.
Constrained diffusion models often fail because training-time optimization uses states from the training distribution, which differ from sampling-time states. The proposed rollout-based fine-tuning aligns training with sampling by differentiating through the fixed noise schedule, improving constraint satisfaction while maintaining sampling quality.
Constrained generative models aim to produce samples that satisfy complex feasibility constraints while remaining faithful to the data distribution. Existing constrained generation methods typically enforce constraints either through training-time optimization or sampling-time correction. Training-time optimization approaches optimize on states induced by the training distribution, which can differ substantially from those encountered during sampling. Sampling-time correction methods instead modify the sampling process at inference, introducing distribution shift and requiring expensive tuning, particularly for few-step sampling. We propose a fine-tuning framework that incorporates constraint guidance obtained through online rollout into the training process, which aligns training with sampling by differentiating through the fixed noise schedule used to numerically integrate the denoising process. This exposes the model to violations that arise along the denoising trajectory and aligns diffusion learning with the sampling process. Experiments across multiple tasks show that our method improves constraint satisfaction while maintaining competitive sampling quality compared to prior methods.