Be like a Goldfish, Don't Memorize! Mitigating Memorization in Generative LLMs
Abstract
A goldfish loss, which excludes a random subset of tokens from the training loss, reduces memorization in large language models with minimal impact on performance.
Large language models can memorize and repeat their training data, causing privacy and copyright risks. To mitigate memorization, we introduce a subtle modification to the next-token training objective that we call the goldfish loss. During training, a randomly sampled subset of tokens are excluded from the loss computation. These dropped tokens are not memorized by the model, which prevents verbatim reproduction of a complete chain of tokens from the training set. We run extensive experiments training billion-scale Llama-2 models, both pre-trained and trained from scratch, and demonstrate significant reductions in extractable memorization with little to no impact on downstream benchmarks.
Community
Models citing this paper 7
Browse 7 models citing this paperDatasets citing this paper 1
Spaces citing this paper 0
No Space linking this paper