Improve Decoding Factuality by Token-wise Cross Layer Entropy of Large Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Wu, Jialiang, Shen, Yi, Liu, Sijia, Tang, Yi, Song, Sen, Wang, Xiaoyi, Cai, Longjun
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910815238488064
author Wu, Jialiang
Shen, Yi
Liu, Sijia
Tang, Yi
Song, Sen
Wang, Xiaoyi
Cai, Longjun
author_facet Wu, Jialiang
Shen, Yi
Liu, Sijia
Tang, Yi
Song, Sen
Wang, Xiaoyi
Cai, Longjun
contents Despite their impressive capacities, Large language models (LLMs) often struggle with the hallucination issue of generating inaccurate or fabricated content even when they possess correct knowledge. In this paper, we extend the exploration of the correlation between hidden-state prediction changes and output factuality into a deeper, token-wise level. Based on the insights , we propose cross-layer Entropy eNhanced Decoding (END), a decoding method that mitigates hallucinations without requiring extra training. END leverages inner probability changes across layers to individually quantify the factual knowledge required for each candidate token, and adjusts the final predicting distribution to prioritize tokens with higher factuality. Experiments on both hallucination and QA benchmarks demonstrate that END significantly enhances the truthfulness and informativeness of generated content while maintaining robust QA accuracy. Moreover, our work provides a deeper perspective on understanding the correlations between inherent knowledge and output factuality.
format Preprint
id arxiv_https___arxiv_org_abs_2502_03199
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Improve Decoding Factuality by Token-wise Cross Layer Entropy of Large Language Models
Wu, Jialiang
Shen, Yi
Liu, Sijia
Tang, Yi
Song, Sen
Wang, Xiaoyi
Cai, Longjun
Computation and Language
Artificial Intelligence
Despite their impressive capacities, Large language models (LLMs) often struggle with the hallucination issue of generating inaccurate or fabricated content even when they possess correct knowledge. In this paper, we extend the exploration of the correlation between hidden-state prediction changes and output factuality into a deeper, token-wise level. Based on the insights , we propose cross-layer Entropy eNhanced Decoding (END), a decoding method that mitigates hallucinations without requiring extra training. END leverages inner probability changes across layers to individually quantify the factual knowledge required for each candidate token, and adjusts the final predicting distribution to prioritize tokens with higher factuality. Experiments on both hallucination and QA benchmarks demonstrate that END significantly enhances the truthfulness and informativeness of generated content while maintaining robust QA accuracy. Moreover, our work provides a deeper perspective on understanding the correlations between inherent knowledge and output factuality.
title Improve Decoding Factuality by Token-wise Cross Layer Entropy of Large Language Models
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2502.03199