差分隐私能减少语言模型在部分任务中的社会偏见，但效果不普适。
How Does Differential Privacy Affect Social Bias in LLMs? A Systematic Evaluation
- 用差分隐私训练模型，限制单条数据影响。
- 在句子评分任务中偏见下降，但其他任务未改善。
- 需多角度评估公平性，不能只看记忆程度。
大规模语言模型在网页规模语料上训练时可能记住敏感数据，带来隐私风险。差分隐私（DP）作为一种原则性框架，可限制单个数据点对训练的影响，但其与模型社会偏见的关系尚不明确。为此，我们系统评估了使用DP-SGD训练的预训练语言模型在四种互补范式下的社会偏见：句子评分、文本补全、表格分类和问答任务。结果发现，差分隐私在句子评分任务中降低了偏见（通过受控似然比较衡量），但该改善无法推广到所有任务。研究揭示了逻辑值层级偏见与输出层级偏见之间的差异。此外，降低记忆并不必然减少不公平性，强调在评估语言模型公平性时必须采用多范式方法。
原文摘要 · Abstract (English)
Large language models (LLMs) trained on web-scale corpora can memorize sensitive training data, posing significant privacy risks. Differential privacy (DP) has emerged as a principled framework that limits the influence of individual data points during training, yet the relationship between differential privacy and social bias in LLMs remains poorly understood. To investigate this, we present a systematic evaluation of social bias in a pretrained LLM trained with DP-SGD, comparing a DP model against non-DP baselines across four complementary paradigms: sentence scoring, text completion, tabular classification, and question answering. We find that DP reduces bias in sentence scoring tasks, where bias is measured through controlled likelihood comparisons, yet this improvement does not generalize across all tasks. Our results reveal a discrepancy between logit-level bias and output-level bias. Moreover, decreasing memorization does not necessarily reduce unfairness, underscoring the importance of multi-paradigm evaluation when assessing fairness in LLMs.
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。