Scalable Multi-Task Low-Rank Model Adaptation

2026-03-02

AI Summary

This work addresses the significant performance degradation of multi-task Low-Rank Adaptation (LoRA) at scale, which stems from parameter misalignment and gradient conflict caused by uniform regularization that disrupts high-singular-value shared knowledge and component-level adaptation that exacerbates task interference. To mitigate these issues, the authors propose mtLoRA, which introduces spectral-aware regularization to preserve critical shared representations, integrates block-level low-rank adaptation—reducing task conflict by 76%—and employs a dimension-specific fine-grained routing mechanism. Evaluated on four benchmarks—DOTA, iNat2018, Dolly-15k, and BBH—mtLoRA achieves an average accuracy gain of 2.3% over state-of-the-art methods while using 47% fewer parameters and requiring 24% less training time.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Large Vision Models
• Natural Language Processing: Safety and Robustness

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Scaling multi-task low-rank adaptation (LoRA) to a large number of tasks induces catastrophic performance degradation, such as an accuracy drop from 88.2% to 2.0% on DOTA when scaling from 5 to 15 tasks. This failure is due to parameter and representation misalignment. We find that existing solutions, like regularization and dynamic routing, fail at scale because they are constrained by a fundamental trade-off: strengthening regularization to reduce inter-task conflict inadvertently suppresses the essential feature discrimination required for effective routing. In this work, we identify two root causes for this trade-off. First, uniform regularization disrupts inter-task knowledge sharing: shared underlying knowledge concentrates in high-SV components (89% alignment on Flanv2->BBH). Uniform regularization forces high-SV components to update in orthogonal directions, directly disrupting the shared knowledge. Second, Conflict Amplification: Applying LoRA at the component-level (e.g., W_q, W_v) amplifies gradient conflicts; we show block-level adaptation reduces this conflict by 76% with only 50% parameters. Based on these insights, we propose mtLoRA, a scalable solution with three novel designs: 1) Spectral-Aware Regularization to selectively orthogonalize low-SV components while preserving high-SV shared knowledge, 2) Block-Level Adaptation to mitigate conflict amplification and largely improve parameter efficiency, and 3) Fine-Grained Routing using dimension-specific weights for superior expressive power. On four large-scale (15-25 tasks) vision (DOTA and iNat2018) and NLP (Dolly-15k and BBH) benchmarks, mtLoRA achieves 91.7%, 81.5%, 44.5% and 38.5% accuracy on DOTA, iNat2018, Dolly-15k and BBH respectively, outperforming the state-of-the-art by 2.3% on average while using 47% fewer parameters and 24% less training time.

Problem

• multi-task learning
• low-rank adaptation
• parameter conflict
• representation misalignment
• scalability

Innovation

• Spectral-Aware Regularization
• Block-Level Adaptation
• Fine-Grained Routing
• Multi-Task LoRA
• Gradient Conflict Mitigation