提出GRU方法，让大模型删记忆时少伤其他能力
GRU: Mitigating the Trade-off between Unlearning and Retention for LLMs
- 用梯度方向调节机制，减少删记忆时的副作用
- 在多个基准上实现更好删记效果，且通用性强
- 适合需要安全合规又不想损失功能的场景
大语言模型（LLM）的去遗忘技术在移除隐私和版权相关响应方面至关重要，有助于其合法安全应用。然而，追求完全去遗忘常以牺牲模型通用功能为代价，导致去遗忘与保留能力之间的显著权衡。为此，本文提出梯度修正去遗忘（GRU）框架，通过调控去遗忘过程中的梯度更新方向，最小化对无关内容的副作用。GRU实现简单、通用性强，在多个主流去遗忘基准测试中均表现出良好的实际效果。
原文摘要 · Abstract (English)
Large language model (LLM) unlearning has demonstrated its essential role in removing privacy and copyright-related responses, crucial for their legal and safe applications. However, the pursuit of complete unlearning often comes with substantial costs due to its compromises in their general functionality, leading to a notorious trade-off between unlearning and retention. It motivates this paper to explore enhanced unlearning schemes that can mitigate this trade-off. Specifically, we propose Gradient Rectified Unlearning (GRU), an improved framework that regulates the directions of gradient updates during the unlearning procedure such that their side impacts on other, unrelated responses can be minimized. GRU is easy and general to implement, demonstrating practical effectiveness across a variety of well-established unlearning benchmarks.
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。