Evolutionary Negative Module Pruning for Better LoRA Merging

TL;DR

This paper introduces ENMP, an evolutionary pruning method that identifies and removes negative LoRA modules to improve the merging of multiple LoRA experts, enhancing multi-task model performance.

Contribution

ENMP is a novel evolutionary search-based pruning technique that effectively excludes detrimental modules before merging, leading to state-of-the-art results in language and vision tasks.

Findings

ENMP consistently improves merging performance across tasks.

It achieves state-of-the-art results in language and vision domains.

The method effectively identifies negative modules that degrade performance.

Abstract

Merging multiple Low-Rank Adaptation (LoRA) experts into a single backbone is a promising approach for efficient multi-task deployment. While existing methods strive to alleviate interference via weight interpolation or subspace alignment, they rest upon the implicit assumption that all LoRA matrices contribute constructively to the merged model. In this paper, we uncover a critical bottleneck in current merging paradigms: the existence of $\textit{negative modules}$ -- specific LoRA layers that inherently degrade global performance upon merging. We propose $\textbf{E}$volutionary $\textbf{N}$egative $\textbf{M}$odule $\textbf{P}$runing ($\textbf{ENMP}$), a plug-and-play LoRA pruning method to locate and exclude these detrimental modules prior to merging. By leveraging an evolutionary search strategy, ENMP effectively navigates the discrete, non-differentiable landscape of module selection to identify optimal pruning configurations. Extensive evaluations demonstrate that ENMP consistently boosts the performance of existing merging algorithms, achieving a new state-of-the-art across both language and vision domains. Code is available at https://github.com/CaoAnda/ENMP-LoRAMerging.