Saved in:
Bibliographic Details
Main Authors: Huang, Wenke, Liang, Jian, Shi, Zekun, Zhu, Didi, Wan, Guancheng, Li, He, Du, Bo, Tao, Dacheng, Ye, Mang
Format: Preprint
Published: 2024
Subjects:
Online Access:https://arxiv.org/abs/2411.10928
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866929594751254528
author Huang, Wenke
Liang, Jian
Shi, Zekun
Zhu, Didi
Wan, Guancheng
Li, He
Du, Bo
Tao, Dacheng
Ye, Mang
author_facet Huang, Wenke
Liang, Jian
Shi, Zekun
Zhu, Didi
Wan, Guancheng
Li, He
Du, Bo
Tao, Dacheng
Ye, Mang
contents Multimodal Large Language Model (MLLM) have demonstrated strong generalization capabilities across diverse distributions and tasks, largely due to extensive pre-training datasets. Fine-tuning MLLM has become a common practice to improve performance on specific downstream tasks. However, during fine-tuning, MLLM often faces the risk of forgetting knowledge acquired during pre-training, which can result in a decline in generalization abilities. To balance the trade-off between generalization and specialization, we propose measuring the parameter importance for both pre-trained and fine-tuning distributions, based on frozen pre-trained weight magnitude and accumulated fine-tuning gradient values. We further apply an importance-aware weight allocation strategy, selectively updating relatively important parameters for downstream tasks. We conduct empirical evaluations on both image captioning and visual question-answering tasks using various MLLM architectures. The comprehensive experimental analysis demonstrates the effectiveness of the proposed solution, highlighting the efficiency of the crucial modules in enhancing downstream specialization performance while mitigating generalization degradation in MLLM Fine-Tuning.
format Preprint
id arxiv_https___arxiv_org_abs_2411_10928
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Learn from Downstream and Be Yourself in Multimodal Large Language Model Fine-Tuning
Huang, Wenke
Liang, Jian
Shi, Zekun
Zhu, Didi
Wan, Guancheng
Li, He
Du, Bo
Tao, Dacheng
Ye, Mang
Computation and Language
Artificial Intelligence
Multimodal Large Language Model (MLLM) have demonstrated strong generalization capabilities across diverse distributions and tasks, largely due to extensive pre-training datasets. Fine-tuning MLLM has become a common practice to improve performance on specific downstream tasks. However, during fine-tuning, MLLM often faces the risk of forgetting knowledge acquired during pre-training, which can result in a decline in generalization abilities. To balance the trade-off between generalization and specialization, we propose measuring the parameter importance for both pre-trained and fine-tuning distributions, based on frozen pre-trained weight magnitude and accumulated fine-tuning gradient values. We further apply an importance-aware weight allocation strategy, selectively updating relatively important parameters for downstream tasks. We conduct empirical evaluations on both image captioning and visual question-answering tasks using various MLLM architectures. The comprehensive experimental analysis demonstrates the effectiveness of the proposed solution, highlighting the efficiency of the crucial modules in enhancing downstream specialization performance while mitigating generalization degradation in MLLM Fine-Tuning.
title Learn from Downstream and Be Yourself in Multimodal Large Language Model Fine-Tuning
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2411.10928