arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2605.08765cs.LGcs.AI
收藏

无法学习者可以撒谎:评估和改进LLM去学习中的诚实性

Unlearners Can Lie: Evaluating and Improving Honesty in LLM Unlearning

Fudan University(复旦大学);Central South University(中南大学);Michigan State University(密歇根州立大学)

Renjie Gu, Jiazhen Du, Yihua Zhang, Sijia Liu

AI总结:

本文评估并改进LLM去学习中的诚实性,提出正式定义并引入评估指标,实验显示现有方法无法满足标准,提出ReVa方法提升去学习效果和诚实性。

摘要:

在大型语言模型(LLM)中,去学习旨在移除有害训练数据同时保持整体效用。然而,我们发现现有方法常产生幻觉、生成异常token序列或行为不一致,引发安全和信任问题。根据先前关于LLM诚实性的文献,此类行为常与不诚实相关。因此,我们研究去学习中的诚实概念,提出去学习诚实的正式定义,包括:(1)保留知识的效用和诚实,(2)确保有效遗忘同时鼓励模型承认其局限性并一致回答遗忘知识相关的问题。为系统评估去学习的诚实性,我们引入涵盖效用、保留集诚实性、遗忘有效性、问答和多项选择题设置中拒绝率和拒绝稳定性的指标。评估9种方法显示所有现有方法均无法满足这些标准。经过实验和理论分析,我们提出ReVa,一种特征随机化去学习模型的表示对齐过程,通过微调以更好地承认遗忘知识。在遗忘集的问答任务中,ReVa在两次交互后实现最高拒绝率,几乎使第二好方法的性能翻倍。值得注意的是,它也提升了保留集的诚实性。我们发布数据和代码在https://github.com/renjiegu。

英文摘要:

Unlearning in large language models (LLMs) aims to remove harmful training data while preserving overall utility. However, we find that existing methods often hallucinate, generate abnormal token sequences, or behave inconsistently, raising safety and trust concerns. According to prior literature on LLM honesty, such behaviors are often associated with dishonesty. This motivates us to investigate the notion of honesty in the context of model unlearning. We propose a formal definition of unlearning honesty, which includes: (1) preserving both utility and honesty on retained knowledge, and (2) ensuring effective forgetting while encouraging the model to acknowledge its limitations and respond consistently to questions related to forgotten knowledge. To systematically evaluate the honesty of unlearning, we introduce a suite of metrics that cover utility, honesty on the retained set, effectiveness of forgetting, rejection rate and refusal stability in Q&A and MCQ settings. Evaluating 9 methods across 3 mainstream families shows that all current methods fail to meet these standards. After experimental and theoretical analyses, we present ReVa, a representation-alignment procedure that fine-tunes feature-randomized unlearned models to better acknowledge forgotten knowledge. On Q&A tasks from the forget set, ReVa achieves the highest rejection rate after two rounds of interaction, nearly doubling the performance of the second-best method. Remarkably, It also improves honesty on the retained set. We release our data and code at https://github.com/renjiegu.

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。

↑