Improving Uncertainty Estimation through Semantically Diverse Language Generation

TL;DR

This paper introduces SDLG, a method that guides large language models to generate semantically diverse alternatives, improving the detection of hallucinations and uncertainty estimation in text generation tasks.

Contribution

The paper presents a novel semantically diverse language generation technique that enhances uncertainty quantification and hallucination detection in large language models.

Findings

SDLG outperforms existing uncertainty estimation methods.

SDLG is computationally efficient.

SDLG improves hallucination detection in question-answering tasks.

Abstract

Large language models (LLMs) can suffer from hallucinations when generating text. These hallucinations impede various applications in society and industry by making LLMs untrustworthy. Current LLMs generate text in an autoregressive fashion by predicting and appending text tokens. When an LLM is uncertain about the semantic meaning of the next tokens to generate, it is likely to start hallucinating. Thus, it has been suggested that predictive uncertainty is one of the main causes of hallucinations. We introduce Semantically Diverse Language Generation (SDLG) to quantify predictive uncertainty in LLMs. SDLG steers the LLM to generate semantically diverse yet likely alternatives for an initially generated text. This approach provides a precise measure of aleatoric semantic uncertainty, detecting whether the initial text is likely to be hallucinated. Experiments on question-answering tasks demonstrate that SDLG consistently outperforms existing methods while being the most computationally efficient, setting a new standard for uncertainty estimation in LLMs.