The Web Can Be Your Oyster for Improving Large Language Models

TL;DR

This paper introduces UNIWEB, a web-augmented large language model that uses adaptive web retrieval and continual knowledge learning to enhance performance on knowledge-intensive tasks, outperforming previous methods.

Contribution

The paper presents a novel web-augmented LLM with adaptive retrieval and a new pretraining task, improving knowledge integration and task performance.

Findings

Significant performance improvements over previous retrieval-augmented models.

Effective use of adaptive search to reduce noise and irrelevant web data.

Successful application across 16 knowledge-intensive tasks.

Abstract

Large language models (LLMs) encode a large amount of world knowledge. However, as such knowledge is frozen at the time of model training, the models become static and limited by the training data at that time. In order to further improve the capacity of LLMs for knowledge-intensive tasks, we consider augmenting LLMs with the large-scale web using search engine. Unlike previous augmentation sources (e.g., Wikipedia data dump), the web provides broader, more comprehensive and constantly updated information. In this paper, we present a web-augmented LLM UNIWEB, which is trained over 16 knowledge-intensive tasks in a unified text-to-text format. Instead of simply using the retrieved contents from web, our approach has made two major improvements. Firstly, we propose an adaptive search engine assisted learning method that can self-evaluate the confidence level of LLM's predictions, and adaptively determine when to refer to the web for more data, which can avoid useless or noisy augmentation from web. Secondly, we design a pretraining task, i.e., continual knowledge learning, based on salient spans prediction, to reduce the discrepancy between the encoded and retrieved knowledge. Experiments on a wide range of knowledge-intensive tasks show that our model significantly outperforms previous retrieval-augmented methods.