Enregistré dans:
Détails bibliographiques
Auteurs principaux: Zhu, Didi, Sun, Zhongyi, Li, Zexi, Shen, Tao, Yan, Ke, Ding, Shouhong, Kuang, Kun, Wu, Chao
Format: Preprint
Publié: 2024
Sujets:
Accès en ligne:https://arxiv.org/abs/2402.12048
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866909111511154688
author Zhu, Didi
Sun, Zhongyi
Li, Zexi
Shen, Tao
Yan, Ke
Ding, Shouhong
Kuang, Kun
Wu, Chao
author_facet Zhu, Didi
Sun, Zhongyi
Li, Zexi
Shen, Tao
Yan, Ke
Ding, Shouhong
Kuang, Kun
Wu, Chao
contents Catastrophic forgetting emerges as a critical challenge when fine-tuning multi-modal large language models (MLLMs), where improving performance on unseen tasks often leads to a significant performance drop on the original tasks. This paper presents a comprehensive analysis of catastrophic forgetting in MLLMs and introduces a post-training adjustment method called Model Tailor. Our method primarily preserves the pre-trained parameters while replacing a small number ($\leq$ 10\%) of fine-tuned parameters, maintaining $\sim$ 99\% effectiveness on original tasks versus pre-training, and achieving $\sim$ 97\% on new tasks compared to standard fine-tuning. Specifically, we derive a sparse mask to identify the "model patch", based on a fusion strategy that integrates salience and sensitivity analysis. Subsequently, a compensation mechanism is introduced to "decorate the patch", enhancing the model's performance on both target and original tasks. Additionally, our method is adaptable to multi-task scenarios. Through extensive experiments on InstructBLIP and LLaVA-1.5 in both image captioning and visual question answering tasks, our approach demonstrates significant task adaptability while preserving inherent pre-trained capabilities.
format Preprint
id arxiv_https___arxiv_org_abs_2402_12048
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Model Tailor: Mitigating Catastrophic Forgetting in Multi-modal Large Language Models
Zhu, Didi
Sun, Zhongyi
Li, Zexi
Shen, Tao
Yan, Ke
Ding, Shouhong
Kuang, Kun
Wu, Chao
Computation and Language
Catastrophic forgetting emerges as a critical challenge when fine-tuning multi-modal large language models (MLLMs), where improving performance on unseen tasks often leads to a significant performance drop on the original tasks. This paper presents a comprehensive analysis of catastrophic forgetting in MLLMs and introduces a post-training adjustment method called Model Tailor. Our method primarily preserves the pre-trained parameters while replacing a small number ($\leq$ 10\%) of fine-tuned parameters, maintaining $\sim$ 99\% effectiveness on original tasks versus pre-training, and achieving $\sim$ 97\% on new tasks compared to standard fine-tuning. Specifically, we derive a sparse mask to identify the "model patch", based on a fusion strategy that integrates salience and sensitivity analysis. Subsequently, a compensation mechanism is introduced to "decorate the patch", enhancing the model's performance on both target and original tasks. Additionally, our method is adaptable to multi-task scenarios. Through extensive experiments on InstructBLIP and LLaVA-1.5 in both image captioning and visual question answering tasks, our approach demonstrates significant task adaptability while preserving inherent pre-trained capabilities.
title Model Tailor: Mitigating Catastrophic Forgetting in Multi-modal Large Language Models
topic Computation and Language
url https://arxiv.org/abs/2402.12048