ResearcharXivNEW

InstantFusion: A Shared Latent Interface for Heterogeneous Image Generators

Dai 2026-10-09

Xuanlang DaiJiazi BuYujie Zhou

Diffusion/flow models have emerged as the de facto paradigms for image generation, giving rise to a diverse family of modern generators with increasingly heterogeneous architectures and representations. Despite their remarkable generation capabilities, the cross-model alignment of latent representations and the consistency of their underlying denoising dynamics remain underexplored. In this work,

Read on arXiv

Data aggregated and editorially reviewed by TrendMing.

Key Contributions

Diffusion/flow models have emerged as the de facto paradigms for image generation, giving rise to a diverse family of modern generators with increasingly heterogeneous architectures and representations.
Despite their remarkable generation capabilities, the cross-model alignment of latent representations and the consistency of their underlying denoising dynamics remain underexplored.
In this work,

Research Themes

AIResearch

Back to AI Research