通过程序记忆持续微调大型语言模型
Continual Fine-Tuning of Large Language Models via Program Memory
AI总结：
本文提出ProCL框架，通过程序记忆结构实现持续学习中的高效微调，提升模型在连续任务中的知识保留与适应能力。
摘要：
本文提出ProCL框架，通过程序记忆结构实现持续学习中的高效微调，提升模型在连续任务中的知识保留与适应能力。参数高效微调（PEFT），特别是低秩适应（LoRA），已成为在有限计算资源下适应大型语言模型（LLMs）的标准方法。然而，在连续设置中，模型依次更新并使用小数据集时，传统LoRA更新难以平衡快速适应与知识保留。现有方法通常将低秩空间视为同质更新区域，缺乏调节短期更新如何随时间累积的机制。我们提出一个具有程序记忆的连续LoRA框架，灵感来自神经科学中的互补学习系统。我们的方法，称为ProCL，将LoRA适配器组织成结构化的程序记忆槽，通过输入条件化的注意力动态检索。这使得能够快速且局部化适应，鼓励相似输入重用共享的适配器区域，同时为未来数据保留未使用的容量。槽位随后与底层适配器结合，后者维护一个分布式表示，逐步在任务间积累知识以平衡可塑性与稳定性。我们的方法完全在LoRA参数化内运行，不增加推理成本。在多样化的基准测试中，实验表明与其他连续LoRA策略相比，本方法在保留性和减少灾难性遗忘方面有所改进。
英文摘要：
Parameter-Efficient Fine-Tuning (PEFT), particularly Low-Rank Adaptation (LoRA), has become a standard approach for adapting Large Language Models (LLMs) under limited compute. However, in continual settings where models are updated sequentially with small datasets, conventional LoRA updates struggle to balance rapid adaptation and knowledge retention. Existing methods typically treat the low-rank space as a homogeneous update region, lacking mechanisms to regulate how short-term updates are consolidated over time. We propose a continual LoRA framework with \textbf{Pro}gram memory, inspired by \textbf{C}omplementary \textbf{L}earning Systems in neuroscience. Our approach, dubbed \textbf{ProCL}, organizes LoRA adapters into structured program memory slots that are dynamically retrieved through input-conditioned attention. This enables rapid and localized adaptation, encouraging similar inputs to reuse shared adapter regions while reserving unused capacity for future data. The slots are then combined with the underlying adapter, which maintains a distributed representation that gradually accumulates knowledge across tasks to balance plasticity and stability. Our method operates entirely within the LoRA parameterization and incurs no additional inference cost. Experiments on diverse benchmarks demonstrate improved retention and reduced catastrophic forgetting over other continual LoRA strategies.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
