Quantifying Memorization Across Neural Language Models

TL;DR

This paper quantifies how large language models memorize training data, revealing that memorization increases with model size, data duplication, and context length, raising privacy and utility concerns as models scale.

Contribution

It introduces three log-linear relationships that measure memorization in LMs and highlights the increasing prevalence of memorization with model scaling.

Findings

Memorization grows with model capacity.

Duplicated data increases memorization.

Longer context prompts lead to more memorization.

Abstract

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.