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.

该论文介绍了多模态大型语言模型（MLLMs）中灾难性遗忘现象的全面分析，并引入了一种名为 Model Tailor 的后训练调整方法。该方法主要保留了预训练参数，同时替换了少量（≤ 10%）的微调参数，相比于预训练，在原始任务上保持了约 99% 的有效性，并在新任务上相比标准微调获得了约 97% 的有效性。此外，该方法适用于多任务场景，并通过在 InstructBLIP 和 LLaVA-1.5 上进行广泛实验证明了显著的任务适应能力，同时保持了内在的预训练能力。