SSR-Merge: Subspace Signal Routing for Training-Free LoRA Merging in Diffusion Models

Low-Rank Adaptation (LoRA) merging can efficiently combine diverse generative capabilities from multiple trained LoRAs for a diffusion model. However, existing LoRA merging techniques often suffer from severe parameter interference, causing destructive collisions in the shared parameter space. To address this, we propose Subspace Signal Routing (SSR), which resolves interference by routing internal signals instead of performing parameter-space merge. Specifically, SSR first constructs a unified subspace by concatenating candidate LoRAs along the rank dimension. Next, SSR employs an inverse correlation matrix to decorrelate mixed signals within this space. Finally, a directional guide matrix steers these purified signals into their respective task-specific subspaces. We provide a rigorous theoretical analysis proving that SSR aligns with the Ordinary Least Squares (OLS) solution, thereby ensuring mathematical optimality. We utilize the additivity of sufficient statistics to design a streaming algorithm. This enables on-the-fly updates that significantly reduce memory overhead and computation time. Extensive experiments validate that SSR significantly outperforms state-of-the-art methods while maintaining comparable efficiency. The source code will be made publicly available.

Many AI systems are customized with small plug-in updates that teach them new abilities, such as recognizing new patterns, generating specific content, or following new instructions. Combining several such updates into one model is useful, but simple merging methods often cause the abilities to interfere with each other. SSR-Merge addresses this by guiding the model’s internal signals to the appropriate update instead of simply averaging the updates together. It requires no retraining and adds no extra cost after merging. Experiments across different tasks show that SSR-Merge preserves multiple learned abilities more reliably, making it easier to build one compact model that combines many customized capabilities.