REALM: Retrieval-Augmented Language Model Pre-Training

TL;DR

REALM introduces a retrieval-augmented pre-training method for language models that enhances knowledge access, interpretability, and modularity, significantly improving open-domain question answering performance.

Contribution

It presents a novel unsupervised pre-training approach for a knowledge retriever integrated with language models, enabling effective retrieval of documents during training and inference.

Findings

Outperforms previous models on open-domain QA benchmarks by 4-16% accuracy

First to pre-train a knowledge retriever in an unsupervised manner using masked language modeling

Provides benefits in interpretability and modularity of language models

Abstract

Language model pre-training has been shown to capture a surprising amount of world knowledge, crucial for NLP tasks such as question answering. However, this knowledge is stored implicitly in the parameters of a neural network, requiring ever-larger networks to cover more facts. To capture knowledge in a more modular and interpretable way, we augment language model pre-training with a latent knowledge retriever, which allows the model to retrieve and attend over documents from a large corpus such as Wikipedia, used during pre-training, fine-tuning and inference. For the first time, we show how to pre-train such a knowledge retriever in an unsupervised manner, using masked language modeling as the learning signal and backpropagating through a retrieval step that considers millions of documents. We demonstrate the effectiveness of Retrieval-Augmented Language Model pre-training (REALM) by fine-tuning on the challenging task of Open-domain Question Answering (Open-QA). We compare against state-of-the-art models for both explicit and implicit knowledge storage on three popular Open-QA benchmarks, and find that we outperform all previous methods by a significant margin (4-16% absolute accuracy), while also providing qualitative benefits such as interpretability and modularity.