通过对比偏好优化提升新闻摘要的事实一致性
Improving Factual Consistency of News Summarization by Contrastive Preference Optimization
AI总结：
针对大型语言模型（LLMs）新闻摘要存在的复杂事实幻觉问题，提出对比偏好优化（CPO）分离其生成忠实与虚假内容的倾向，并结合探测式训练增强区分能力，显著提升了摘要可靠性。
摘要：
尽管大型语言模型（LLMs）在新闻摘要领域取得了近期进展，但它们生成的摘要往往与原始文章存在事实不一致，这在文本生成中被称为“幻觉”。与BART、T5等之前的小型模型不同，当前LLMs犯的低级错误更少，但会出现更复杂的错误，例如强加因果关系、添加虚假细节、过度概括等。这些幻觉难以通过传统方法检测，给提升文本摘要的事实一致性带来了巨大挑战。本文提出对比偏好优化（CPO）来分离LLMs生成忠实内容和虚假内容的倾向。此外，我们采用基于探测的特定训练方法来提升其区分这两种倾向的能力。通过这种方式，LLMs能更准确地执行指令，并增强对幻觉的感知。实验结果表明，CPO显著提升了基于LLMs的摘要可靠性。
英文摘要：
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.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
