How Decoding Strategies Affect the Verifiability of Generated Text

TL;DR

This paper examines how different decoding strategies influence the factual verifiability of text generated by language models, revealing a tradeoff between factuality and repetitiveness, and proposing a new decoding method.

Contribution

It introduces a new decoding strategy that balances verifiability and diversity, improving the factual accuracy of generated text compared to existing methods.

Findings

Decoding strategies significantly impact text verifiability.

Top-k and nucleus sampling reduce repetition but lower factuality.

A new decoding method improves both verifiability and diversity.

Abstract

Recent progress in pre-trained language models led to systems that are able to generate text of an increasingly high quality. While several works have investigated the fluency and grammatical correctness of such models, it is still unclear to which extent the generated text is consistent with factual world knowledge. Here, we go beyond fluency and also investigate the verifiability of text generated by state-of-the-art pre-trained language models. A generated sentence is verifiable if it can be corroborated or disproved by Wikipedia, and we find that the verifiability of generated text strongly depends on the decoding strategy. In particular, we discover a tradeoff between factuality (i.e., the ability of generating Wikipedia corroborated text) and repetitiveness. While decoding strategies such as top-k and nucleus sampling lead to less repetitive generations, they also produce less verifiable text. Based on these finding, we introduce a simple and effective decoding strategy which, in comparison to previously used decoding strategies, produces less repetitive and more verifiable text.