提出Gated LoRA机制，解决大模型持续学习中的遗忘问题。
Gated Integration of Low-Rank Adaptation for Continual Learning of Large Language Models
- 为每项新任务新增LoRA分支，并用门控模块融合新旧分支。
- 门控机制显著降低新任务对旧任务的影响，减少遗忘。
- 在多个基准上优于现有方法，适合需要持续更新的大模型场景。
持续学习（CL）要求模型顺序学习多个任务，对大语言模型（LLMs）至关重要。近年来，低秩适配（LoRA）作为最具代表性的参数高效微调（PEFT）方法，在LLM的持续学习中受到广泛关注。然而，现有基于LoRA的持续学习方法通常为每个新任务扩展新的LoRA分支，并强制新旧分支对旧任务产生同等影响，可能导致遗忘。本文提出一种新方法——门控集成低秩适配（GainLoRA），为每个新任务扩展新的LoRA分支，并引入门控模块融合新旧分支。此外，GainLoRA利用新门控模块最小化新LoRA分支对旧任务的影响，有效缓解遗忘并提升整体性能。在持续学习基准上的实验结果表明，GainLoRA优于现有最先进方法。
原文摘要 · Abstract (English)
Continual learning (CL), which requires the model to learn multiple tasks sequentially, is crucial for large language models (LLMs). Recently, low-rank adaptation~(LoRA), one of the most representative parameter-efficient fine-tuning (PEFT) methods, has gained increasing attention in CL of LLMs. However, most existing CL methods based on LoRA typically expand a new LoRA branch to learn each new task and force the new and old LoRA branches to influence old tasks equally, potentially leading to forgetting. In this work, we propose a new method, called gated integration of low-rank adaptation (GainLoRA), for CL of LLMs. GainLoRA expands a new LoRA branch for each new task and introduces gating modules to integrate the new and old LoRA branches. Furthermore, GainLoRA leverages the new gating module to minimize the influence from the new LoRA branch to old tasks, effectively mitigating forgetting and improving the model's overall performance. Experimental results on CL benchmarks demonstrate that GainLoRA outperforms existing state-of-the-art methods.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。