arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2202.07646cs.LGcs.CL
收藏

跨神经语言模型的量化语义化

Quantifying Memorization Across Neural Language Models

Google Research(谷歌研究院);University of Pennsylvania(宾夕法尼亚大学);Cornell University(康奈尔大学)

Nicholas Carlini, Daphne Ippolito, Matthew Jagielski, Katherine Lee, Florian Tramer, Chiyuan Zhang

摘要:

大语言模型(LM)已经被证明可以记忆部分训练数据,并且当适当提示时,它们将逐字发出记忆的训练数据。这是不可取的,因为记忆侵犯隐私(暴露用户数据),降低效用(重复的易于记忆的文本通常质量较低),并损害公平性(一些文本被记住而不是其他文本)。我们描述了三个对数线性关系,量化的程度LM发射记忆训练数据。随着我们增加(1)模型的容量,(2)示例被复制的次数,以及(3)用于提示模型的上下文标记的数量,验证显著增长。令人惊讶的是,我们发现情况变得更加复杂时,概括这些结果跨模型的家庭。总的来说,我们发现LM中的记忆比以前认为的更普遍,并且随着模型的继续扩展,至少在没有积极缓解的情况下,可能会变得更糟。

英文摘要:

Large language models (LMs) have been shown to memorize parts of their training data, and when prompted appropriately, they will emit the memorized training data verbatim. This is undesirable because memorization violates privacy (exposing user data), degrades utility (repeated easy-to-memorize text is often low quality), and hurts fairness (some texts are memorized over others). We describe three log-linear relationships that quantify the degree to which LMs emit memorized training data. Memorization significantly grows as we increase (1) the capacity of a model, (2) the number of times an example has been duplicated, and (3) the number of tokens of context used to prompt the model. Surprisingly, we find the situation becomes more complicated when generalizing these results across model families. On the whole, we find that memorization in LMs is more prevalent than previously believed and will likely get worse as models continues to scale, at least without active mitigations.

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。

↑