Improving Factual Consistency of News Summarization by Contrastive Preference Optimization

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Feng, Huawen, Fan, Yan, Liu, Xiong, Lin, Ting-En, Yao, Zekun, Wu, Yuchuan, Huang, Fei, Li, Yongbin, Ma, Qianli
Format: Preprint
Published: 2023
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912231667531776
author Feng, Huawen
Fan, Yan
Liu, Xiong
Lin, Ting-En
Yao, Zekun
Wu, Yuchuan
Huang, Fei
Li, Yongbin
Ma, Qianli
author_facet Feng, Huawen
Fan, Yan
Liu, Xiong
Lin, Ting-En
Yao, Zekun
Wu, Yuchuan
Huang, Fei
Li, Yongbin
Ma, Qianli
contents Despite the recent progress in news summarization made by large language models (LLMs), they often generate summaries that are factually inconsistent with original articles, known as "hallucinations" in text generation. Unlike previous small models (e.g., BART, T5), current LLMs make fewer silly mistakes but more sophisticated ones, such as imposing cause and effect, adding false details, overgeneralizing, etc. These hallucinations are challenging to detect through traditional methods, which poses great challenges for improving the factual consistency of text summarization. In this paper, we propose Contrastive Preference Optimization (CPO) to disentangle the LLMs' propensities to generate faithful and fake content. Furthermore, we adopt a probing-based specific training method to improve their capacity of distinguishing two types of propensities. In this way, LLMs can execute the instructions more accurately and have enhanced perception of hallucinations. Experimental results show that CPO significantly improves the reliability of summarization based on LLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2310_19347
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle Improving Factual Consistency of News Summarization by Contrastive Preference Optimization
Feng, Huawen
Fan, Yan
Liu, Xiong
Lin, Ting-En
Yao, Zekun
Wu, Yuchuan
Huang, Fei
Li, Yongbin
Ma, Qianli
Computation and Language
Artificial Intelligence
Despite the recent progress in news summarization made by large language models (LLMs), they often generate summaries that are factually inconsistent with original articles, known as "hallucinations" in text generation. Unlike previous small models (e.g., BART, T5), current LLMs make fewer silly mistakes but more sophisticated ones, such as imposing cause and effect, adding false details, overgeneralizing, etc. These hallucinations are challenging to detect through traditional methods, which poses great challenges for improving the factual consistency of text summarization. In this paper, we propose Contrastive Preference Optimization (CPO) to disentangle the LLMs' propensities to generate faithful and fake content. Furthermore, we adopt a probing-based specific training method to improve their capacity of distinguishing two types of propensities. In this way, LLMs can execute the instructions more accurately and have enhanced perception of hallucinations. Experimental results show that CPO significantly improves the reliability of summarization based on LLMs.
title Improving Factual Consistency of News Summarization by Contrastive Preference Optimization
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2310.19347