English

Diffusion Soup: Model Merging for Text-to-Image Diffusion Models

Computer Vision and Pattern Recognition 2024-06-13 v1 Artificial Intelligence Cryptography and Security Machine Learning

Abstract

We present Diffusion Soup, a compartmentalization method for Text-to-Image Generation that averages the weights of diffusion models trained on sharded data. By construction, our approach enables training-free continual learning and unlearning with no additional memory or inference costs, since models corresponding to data shards can be added or removed by re-averaging. We show that Diffusion Soup samples from a point in weight space that approximates the geometric mean of the distributions of constituent datasets, which offers anti-memorization guarantees and enables zero-shot style mixing. Empirically, Diffusion Soup outperforms a paragon model trained on the union of all data shards and achieves a 30% improvement in Image Reward (.34 \to .44) on domain sharded data, and a 59% improvement in IR (.37 \to .59) on aesthetic data. In both cases, souping also prevails in TIFA score (respectively, 85.5 \to 86.5 and 85.6 \to 86.8). We demonstrate robust unlearning -- removing any individual domain shard only lowers performance by 1% in IR (.45 \to .44) -- and validate our theoretical insights on anti-memorization using real data. Finally, we showcase Diffusion Soup's ability to blend the distinct styles of models finetuned on different shards, resulting in the zero-shot generation of hybrid styles.

Keywords

Cite

@article{arxiv.2406.08431,
  title  = {Diffusion Soup: Model Merging for Text-to-Image Diffusion Models},
  author = {Benjamin Biggs and Arjun Seshadri and Yang Zou and Achin Jain and Aditya Golatkar and Yusheng Xie and Alessandro Achille and Ashwin Swaminathan and Stefano Soatto},
  journal= {arXiv preprint arXiv:2406.08431},
  year   = {2024}
}
R2 v1 2026-06-28T17:03:27.283Z