Generalization through Memorization: Nearest Neighbor Language Models

TL;DR

This paper presents $k$NN-LMs, a method that enhances pre-trained language models by interpolating with nearest neighbor models, leading to state-of-the-art perplexity and improved domain adaptation without additional training.

Contribution

The introduction of $k$NN-LMs that combine neural language models with nearest neighbor retrieval, enabling better performance and domain adaptation without extra training.

Findings

Achieved new state-of-the-art perplexity of 15.79 on Wikitext-103

Improved language modeling performance without additional training

Effective in predicting rare and factual patterns

Abstract

We introduce $k$NN-LMs, which extend a pre-trained neural language model (LM) by linearly interpolating it with a $k$-nearest neighbors ($k$NN) model. The nearest neighbors are computed according to distance in the pre-trained LM embedding space, and can be drawn from any text collection, including the original LM training data. Applying this augmentation to a strong Wikitext-103 LM, with neighbors drawn from the original training set, our $k$NN-LM achieves a new state-of-the-art perplexity of 15.79 - a 2.9 point improvement with no additional training. We also show that this approach has implications for efficiently scaling up to larger training sets and allows for effective domain adaptation, by simply varying the nearest neighbor datastore, again without further training. Qualitatively, the model is particularly helpful in predicting rare patterns, such as factual knowledge. Together, these results strongly suggest that learning similarity between sequences of text is easier than predicting the next word, and that nearest neighbor search is an effective approach for language modeling in the long tail.