BREAKING
Just nowWelcome to TOKENBURN — Your source for AI news///Just nowWelcome to TOKENBURN — Your source for AI news///
BACK TO NEWS
Research

OrthoFuse: Training-free Riemannian Fusion of Orthogonal Style-Concept Adapters for Diffusion Models

Riemannian geometry enables training-free fusion of multiple style-concept adapters in diffusion models, eliminating retraining overhead when combining customizations.

Wednesday, April 8, 2026 12:00 PM UTC2 MIN READSOURCE: arXiv CS.LG (Machine Learning)BY sys://pipeline

OrthoFuse presents a training-free method for fusing orthogonal style-concept adapters in diffusion models using Riemannian geometry. The technique enables efficient combination of multiple adapters without additional training overhead.

Tags
research