Graceful Forgetting in Generative Language Models

ACL ARR 2025 February Submission1995 Authors

Abstract: Recently, the pretrain-finetune paradigm has become a cornerstone in various deep learning areas. While generally, the pre-trained model would provide both effectiveness and efficiency to downstream fine-tuning, studies have shown that not all knowledge acquired during pre-training is beneficial. Some of the knowledge may actually bring detrimental effects. To address this negative transfer problem, graceful forgetting has emerged as a promising approach. The core principle of graceful forgetting is to enhance the learning plasticity of the target task by selectively discarding knowledge from irrelevant tasks. However, this approach remains underexplored in the context of generative language models, and it is often challenging to migrate existing graceful forgetting algorithms to these models due to architecture incompatibility. To bridge this gap, in this paper we propose a novel framework, Learning With Forgetting (LWF), to achieve graceful forgetting in generative language models. With Fisher Information Matrix weighting the intended parameter updates, LWF computes forgetting confidence to evaluate self-generated knowledge regarding the forgetting task, and consequently, knowledge with high confidence is periodically unlearned during fine-tuning. We evaluate our framework on domain-specific question-answering tasks, demonstrating that, although determining the inter-task interaction mechanisms is still highly tricky, graceful forgetting can indeed lead to improved fine-tuning.

Paper Type: Long

Research Area: Language Modeling

Research Area Keywords: fine-tuning, generative models

Contribution Types: Model analysis & interpretability, NLP engineering experiment, Data analysis

Languages Studied: English

Submission Number: 1995