CorpusBrain++: A Continual Generative Pre-Training Framework for Knowledge-Intensive Language Tasks

TL;DR

CorpusBrain++ is a new continual pre-training framework designed to improve knowledge-intensive language tasks by effectively handling dynamic document updates, addressing the limitations of static retrieval models like CorpusBrain.

Contribution

It introduces the CDL task and KILT++ benchmark, and proposes CorpusBrain++, a continual learning approach that mitigates catastrophic forgetting in dynamic retrieval scenarios.

Findings

CorpusBrain++ outperforms traditional IR methods in dynamic settings.

CorpusBrain++ significantly reduces catastrophic forgetting.

Empirical results show improved retrieval performance and efficiency.

Abstract

Knowledge-intensive language tasks (KILTs) typically require retrieving relevant documents from trustworthy corpora, e.g., Wikipedia, to produce specific answers. Very recently, a pre-trained generative retrieval model for KILTs, named CorpusBrain, was proposed and reached new state-of-the-art retrieval performance. However, most existing research on KILTs, including CorpusBrain, has predominantly focused on a static document collection, overlooking the dynamic nature of real-world scenarios, where new documents are continuously being incorporated into the source corpus. To address this gap, it is crucial to explore the capability of retrieval models to effectively handle the dynamic retrieval scenario inherent in KILTs. In this work, we first introduce the continual document learning (CDL) task for KILTs and build a novel benchmark dataset named KILT++ based on the original KILT dataset for evaluation. Then, we conduct a comprehensive study over the use of pre-trained CorpusBrain on KILT++. Unlike the promising results in the stationary scenario, CorpusBrain is prone to catastrophic forgetting in the dynamic scenario, hence hampering the retrieval performance. To alleviate this issue, we propose CorpusBrain++, a continual generative pre-training framework. Empirical results demonstrate the significant effectiveness and remarkable efficiency of CorpusBrain++ in comparison to both traditional and generative IR methods.