Entropy Guided Extrapolative Decoding to Improve Factuality in Large Language Models

TL;DR

This paper introduces an entropy-guided extrapolative decoding method that enhances factual accuracy in large language models by leveraging layer-wise information and extrapolating token probabilities beyond the last layer.

Contribution

It proposes a novel entropy-guided layer selection and probability extrapolation technique that improves factuality in LLM outputs, surpassing current state-of-the-art methods.

Findings

Outperforms existing methods on multiple datasets.

Layer-wise entropy guides better token selection.

Extrapolation improves factuality accuracy.

Abstract

Large language models (LLMs) exhibit impressive natural language capabilities but suffer from hallucination -- generating content ungrounded in the realities of training data. Recent work has focused on decoding techniques to improve factuality during inference by leveraging LLMs' hierarchical representation of factual knowledge, manipulating the predicted distributions at inference time. Current state-of-the-art approaches refine decoding by contrasting early-exit distributions from a lower layer with the final layer to exploit information related to factuality within the model forward procedure. However, such methods often assume the final layer is the most reliable and the lower layer selection process depends on it. In this work, we first propose extrapolation of critical token probabilities beyond the last layer for more accurate contrasting. We additionally employ layer-wise entropy-guided lower layer selection, decoupling the selection process from the final layer. Experiments demonstrate strong performance - surpassing state-of-the-art on multiple different datasets by large margins. Analyses show different kinds of prompts respond to different selection strategies.