REPLUG: Retrieval-Augmented Black-Box Language Models

TL;DR

REPLUG is a retrieval-augmented framework that enhances black-box language models by prepending retrieved documents, improving their performance without retraining the models themselves.

Contribution

It introduces a simple, adaptable retrieval augmentation method that treats language models as black boxes and uses the model to supervise the retriever.

Findings

GPT-3 performance improved by 6.3% with REPLUG.

Codex's five-shot MMLU accuracy increased by 5.1%.

The method is easily applicable to existing models without retraining.

Abstract

We introduce REPLUG, a retrieval-augmented language modeling framework that treats the language model (LM) as a black box and augments it with a tuneable retrieval model. Unlike prior retrieval-augmented LMs that train language models with special cross attention mechanisms to encode the retrieved text, REPLUG simply prepends retrieved documents to the input for the frozen black-box LM. This simple design can be easily applied to any existing retrieval and language models. Furthermore, we show that the LM can be used to supervise the retrieval model, which can then find documents that help the LM make better predictions. Our experiments demonstrate that REPLUG with the tuned retriever significantly improves the performance of GPT-3 (175B) on language modeling by 6.3%, as well as the performance of Codex on five-shot MMLU by 5.1%.