KILM: Knowledge Injection into Encoder-Decoder Language Models

TL;DR

KILM introduces a method to inject entity-related knowledge into encoder-decoder language models through continued pre-training, enhancing knowledge retention and reducing hallucinations without changing model architecture.

Contribution

The paper presents a novel knowledge injection technique called KILM that improves knowledge retention in PLMs without architectural modifications or additional parameters.

Findings

Enhanced knowledge retention and reduced hallucinations in models.

Improved zero-shot performance on knowledge-intensive tasks.

Outperforms larger models in entity disambiguation.

Abstract

Large pre-trained language models (PLMs) have been shown to retain implicit knowledge within their parameters. To enhance this implicit knowledge, we propose Knowledge Injection into Language Models (KILM), a novel approach that injects entity-related knowledge into encoder-decoder PLMs, via a generative knowledge infilling objective through continued pre-training. This is done without architectural modifications to the PLMs or adding additional parameters. Experimental results over a suite of knowledge-intensive tasks spanning numerous datasets show that KILM enables models to retain more knowledge and hallucinate less, while preserving their original performance on general NLU and NLG tasks. KILM also demonstrates improved zero-shot performances on tasks such as entity disambiguation, outperforming state-of-the-art models having 30x more parameters.