Disentangling Task Conflicts in Multi-Task LoRA via Orthogonal Gradient Projection

TL;DR

This work tackles gradient conflicts in multi-task learning when using LoRA adapters by introducing Ortho-LoRA, an optimization strategy that enforces orthogonality between task gradients within the LoRA subspace. By applying structure-aware, pairwise gradient projections to the low-rank matrices $\mathbf{A}$ and $\mathbf{B}$, the method mitigates negative transfer while preserving shared representations. Empirical results on GLUE with a RoBERTa-base backbone show Ortho-LoRA markedly reduces interference, achieving performance close to independent single-task adapters with negligible backbone overhead. The approach offers a practical, parameter-efficient route to robust multi-task deployment of large language models and lays groundwork for extending gradient projection to other PEFT paradigms.

Abstract

Multi-Task Learning (MTL) combined with Low-Rank Adaptation (LoRA) has emerged as a promising direction for parameter-efficient deployment of Large Language Models (LLMs). By sharing a single adapter across multiple tasks, one can significantly reduce storage overhead. However, this approach suffers from negative transfer, where conflicting gradient updates from distinct tasks degrade the performance of individual tasks compared to single-task fine-tuning. This problem is exacerbated in LoRA due to the low-rank constraint, which limits the optimization landscape's capacity to accommodate diverse task requirements. In this paper, we propose Ortho-LoRA, a gradient projection method specifically tailored for the bipartite structure of LoRA. Ortho-LoRA dynamically projects conflicting task gradients onto the orthogonal complement of each other within the intrinsic LoRA subspace. Extensive experiments on the GLUE benchmark demonstrate that Ortho-LoRA effectively mitigates task interference, outperforming standard joint training and recovering 95\% of the performance gap between multi-task and single-task baselines with negligible computational overhead.