Controlled Low-Rank Adaptation with Subspace Regularization for Continued Training on Large Language Models
This addresses the problem of performance decline in previous tasks for users of large language models during continual learning, representing an incremental improvement in parameter-efficient fine-tuning methods.
The paper tackles catastrophic forgetting in large language models during adaptation to new tasks by proposing Controlled LoRA (CLoRA), a subspace regularization method that imposes constraints on the updating matrix's null space, resulting in effective mitigation of forgetting while maintaining model capacity.
Large language models (LLMs) exhibit remarkable capabilities in natural language processing but face catastrophic forgetting when learning new tasks, where adaptation to a new domain leads to a substantial decline in performance on previous tasks. In this paper, we propose Controlled LoRA (CLoRA), a sub-space regularization method on LoRA structure. Aiming to reduce the scale of output change while introduce minimal constraint on model capacity, CLoRA imposes constraint on the direction of updating matrix's null space. Experimental results on one-stage LLM finetuning tasks and continual learning settings highlight the superority of CLoRA as a effective parameter efficient finetuning method with catastrophic forgetting mitigating.Further investigation for model parameters indicates that CLoRA effectively balances the trade-off between model capacity and degree of forgetting.