Fine-tuning of diffusion models via stochastic control: entropy regularization and beyond
This work offers a theoretical foundation for fine-tuning diffusion models, which is incremental but important for researchers in generative AI.
This paper provides a rigorous mathematical treatment of entropy-regularized fine-tuning for diffusion models using stochastic control to prevent reward collapse, and extends the approach to general f-divergence regularizers with validation on Stable Diffusion v1.5.
This paper aims to develop and provide a rigorous treatment to the problem of entropy regularized fine-tuning in the context of continuous-time diffusion models, which was recently proposed by Uehara et al. (arXiv:2402.15194, 2024). The idea is to use stochastic control for sample generation, where the entropy regularizer is introduced to mitigate reward collapse. We also show how the analysis can be extended to fine-tuning with a general $f$-divergence regularizer. Numerical experiments on large-scale text-to-image models--Stable Diffusion v1.5 are conducted to validate our approach.