Saved in:
Bibliographic Details
Main Authors: Wei, Xuyang, Tian, Chunlin, Li, Li
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2502.20035
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929734257999872
author Wei, Xuyang
Tian, Chunlin
Li, Li
author_facet Wei, Xuyang
Tian, Chunlin
Li, Li
contents Effective instruction fine-tuning on diverse image-text datasets is crucial for developing a versatile Multimodal Large Language Model (MLLM), where dataset composition dictates the model's adaptability across multimodal tasks. However, complex datasets often contain inherent conflicts -- stemming from modality-specific optimization objectives -- and latent commonalities that enable cross-task transfer, which most existing approaches handle separately. To bridge this gap, we introduce AsymLoRA, a parameter-efficient tuning framework that unifies knowledge modularization and cross-modal coordination via asymmetric LoRA: task-specific low-rank projections (matrix B) that preserve distinct adaptation pathways for conflicting objectives, and a shared projection (matrix A) that consolidates cross-modal commonalities. Extensive evaluations demonstrate that AsymLoRA consistently surpasses both vanilla LoRA, which captures only commonalities, and LoRA-MoE, which focuses solely on conflicts, achieving superior model performance and system efficiency across diverse benchmarks.\href{Code}{https://github.com/Clin0212/HydraLoRA/blob/main/MLLM-HydraLoRA/README.md}.
format Preprint
id arxiv_https___arxiv_org_abs_2502_20035
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle AsymLoRA: Harmonizing Data Conflicts and Commonalities in MLLMs
Wei, Xuyang
Tian, Chunlin
Li, Li
Computer Vision and Pattern Recognition
Effective instruction fine-tuning on diverse image-text datasets is crucial for developing a versatile Multimodal Large Language Model (MLLM), where dataset composition dictates the model's adaptability across multimodal tasks. However, complex datasets often contain inherent conflicts -- stemming from modality-specific optimization objectives -- and latent commonalities that enable cross-task transfer, which most existing approaches handle separately. To bridge this gap, we introduce AsymLoRA, a parameter-efficient tuning framework that unifies knowledge modularization and cross-modal coordination via asymmetric LoRA: task-specific low-rank projections (matrix B) that preserve distinct adaptation pathways for conflicting objectives, and a shared projection (matrix A) that consolidates cross-modal commonalities. Extensive evaluations demonstrate that AsymLoRA consistently surpasses both vanilla LoRA, which captures only commonalities, and LoRA-MoE, which focuses solely on conflicts, achieving superior model performance and system efficiency across diverse benchmarks.\href{Code}{https://github.com/Clin0212/HydraLoRA/blob/main/MLLM-HydraLoRA/README.md}.
title AsymLoRA: Harmonizing Data Conflicts and Commonalities in MLLMs
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2502.20035