ReLearn: Unlearning via Learning for Large Language Models
This work addresses the challenge of effective unlearning for large language models, which is crucial for applications requiring data privacy and model safety, though it appears incremental by improving upon existing methods.
The paper tackles the problem of unlearning in large language models, where existing methods degrade performance and coherence; it proposes ReLearn, a data augmentation and fine-tuning pipeline that achieves targeted forgetting while preserving high-quality output, as demonstrated through experiments.
Current unlearning methods for large language models usually rely on reverse optimization to reduce target token probabilities. However, this paradigm disrupts the subsequent tokens prediction, degrading model performance and linguistic coherence. Moreover, existing evaluation metrics overemphasize contextual forgetting while inadequately assessing response fluency and relevance. To address these challenges, we propose ReLearn, a data augmentation and fine-tuning pipeline for effective unlearning, along with a comprehensive evaluation framework. This framework introduces Knowledge Forgetting Rate (KFR) and Knowledge Retention Rate (KRR) to measure knowledge-level preservation, and Linguistic Score (LS) to evaluate generation quality. Our experiments show that ReLearn successfully achieves targeted forgetting while preserving high-quality output. Through mechanistic analysis, we further demonstrate how reverse optimization disrupts coherent text generation, while ReLearn preserves this essential capability. Code is available at https://github.com/zjunlp/unlearn.