Decom-Renorm-Merge: Model Merging on the Right Space Improves Multitasking

TL;DR

Decom-Renorm-Merge (DRM) introduces a novel SVD-based approach to align and merge neural network models in a shared space, improving multitasking capabilities across various architectures.

Contribution

The paper proposes DRM, a new method that uses Singular Value Decomposition to enable effective model merging by aligning weight matrices in a joint space, outperforming existing techniques.

Findings

DRM outperforms state-of-the-art merging methods.

Renormalization is key to creating a robust joint space.

Effective across various model architectures and sizes.

Abstract

In the era of large-scale training, model merging has evolved into a tool for creating multitasking models efficiently. It enables the knowledge of models to be fused, without the need for heavy computation as required in traditional multitask learning. Existing merging methods often assume that entries at identical positions in weight matrices serve the same function, enabling straightforward entry-wise comparison and merging. However, this assumption overlooks the complexity of finetuned neural networks, where neurons may develop distinct feature compositions, making direct entry-wise merging problematic. We present Decom-Renorm-Merge (DRM), a simple yet effective approach that leverages Singular Value Decomposition to decompose and coordinate weight matrices into an aligned joint space, where entry-wise merging becomes possible. We showcase the effectiveness of DRM across various settings ranging from smaller encoder-based such as ViT and DeBERTa, encoder-decoder-based such as T5, and larger decoder-based such as Llama3.1-8B. Our experimental results show that DRM outperforms several state-of-the-art merging techniques across full finetuning and low-rank adaptation settings. Moreover, our analysis reveals renormalization as the crucial component for creating a robust and even joint space for merging, significantly contributing to the method's performance.