DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Chuang, Yung-Sung, Xie, Yujia, Luo, Hongyin, Kim, Yoon, Glass, James, He, Pengcheng
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910360203689984
author Chuang, Yung-Sung
Xie, Yujia
Luo, Hongyin
Kim, Yoon
Glass, James
He, Pengcheng
author_facet Chuang, Yung-Sung
Xie, Yujia
Luo, Hongyin
Kim, Yoon
Glass, James
He, Pengcheng
contents Despite their impressive capabilities, large language models (LLMs) are prone to hallucinations, i.e., generating content that deviates from facts seen during pretraining. We propose a simple decoding strategy for reducing hallucinations with pretrained LLMs that does not require conditioning on retrieved external knowledge nor additional fine-tuning. Our approach obtains the next-token distribution by contrasting the differences in logits obtained from projecting the later layers versus earlier layers to the vocabulary space, exploiting the fact that factual knowledge in an LLMs has generally been shown to be localized to particular transformer layers. We find that this Decoding by Contrasting Layers (DoLa) approach is able to better surface factual knowledge and reduce the generation of incorrect facts. DoLa consistently improves the truthfulness across multiple choices tasks and open-ended generation tasks, for example improving the performance of LLaMA family models on TruthfulQA by 12-17% absolute points, demonstrating its potential in making LLMs reliably generate truthful facts.
format Preprint
id arxiv_https___arxiv_org_abs_2309_03883
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Models
Chuang, Yung-Sung
Xie, Yujia
Luo, Hongyin
Kim, Yoon
Glass, James
He, Pengcheng
Computation and Language
Artificial Intelligence
Machine Learning
Despite their impressive capabilities, large language models (LLMs) are prone to hallucinations, i.e., generating content that deviates from facts seen during pretraining. We propose a simple decoding strategy for reducing hallucinations with pretrained LLMs that does not require conditioning on retrieved external knowledge nor additional fine-tuning. Our approach obtains the next-token distribution by contrasting the differences in logits obtained from projecting the later layers versus earlier layers to the vocabulary space, exploiting the fact that factual knowledge in an LLMs has generally been shown to be localized to particular transformer layers. We find that this Decoding by Contrasting Layers (DoLa) approach is able to better surface factual knowledge and reduce the generation of incorrect facts. DoLa consistently improves the truthfulness across multiple choices tasks and open-ended generation tasks, for example improving the performance of LLaMA family models on TruthfulQA by 12-17% absolute points, demonstrating its potential in making LLMs reliably generate truthful facts.
title DoLa: Decoding by Contrasting Layers Improves Factuality in Large Language Models
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2309.03883