Group Diffusion: Enhancing Image Generation by Unlocking Cross-Sample Collaboration

TL;DR

This paper introduces Group Diffusion, a novel method that enables collaborative image generation by sharing attention across multiple samples during inference, leading to improved quality and new insights into cross-sample interactions.

Contribution

We propose Group Diffusion, which unlocks cross-sample attention in diffusion models, significantly enhancing image quality and revealing the potential of collaborative inference.

Findings

Larger group sizes improve generation quality.

Cross-sample attention correlates with FID improvements.

Achieves up to 32.2% FID reduction on ImageNet-256x256.

Abstract

In this work, we explore an untapped signal in diffusion model inference. While all previous methods generate images independently at inference, we instead ask if samples can be generated collaboratively. We propose Group Diffusion, unlocking the attention mechanism to be shared across images, rather than limited to just the patches within an image. This enables images to be jointly denoised at inference time, learning both intra and inter-image correspondence. We observe a clear scaling effect - larger group sizes yield stronger cross-sample attention and better generation quality. Furthermore, we introduce a qualitative measure to capture this behavior and show that its strength closely correlates with FID. Built on standard diffusion transformers, our GroupDiff achieves up to 32.2% FID improvement on ImageNet-256x256. Our work reveals cross-sample inference as an effective, previously unexplored mechanism for generative modeling.