Improving Factual Consistency of News Summarization by Contrastive Preference Optimization

TL;DR

This paper introduces Contrastive Preference Optimization (CPO), a novel training method that enhances large language models' ability to generate factually consistent news summaries by better distinguishing faithful content from hallucinations.

Contribution

The paper proposes CPO, a new contrastive learning approach, and a probing-based training method to improve LLMs' factual accuracy in news summarization.

Findings

CPO significantly improves factual consistency in LLM-generated summaries.

Enhanced perception of hallucinations leads to more reliable summarization.

Experimental results demonstrate the effectiveness of the proposed methods.

Abstract

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.