从1到5的评分尺度：量化忠实度评估中的幻觉
On A Scale From 1 to 5: Quantifying Hallucination in Faithfulness Evaluation
AI总结：
本文研究引导式NLG中的自动化忠实度评估，提出基于评分标准模板和LLM的量化打分方法，并开发合成不忠实数据及幻觉百分比量化启发式方法，在4个旅游领域数据集上验证了GPT-4的准确判断能力及合成数据微调NLI模型的有效性。
摘要：
幻觉一直是自然语言生成（NLG）领域的热门话题。在实际应用中，不忠实的内容可能导致数据质量下降或终端用户信任丧失。因此，在将NLG用于生产环境之前进行事实核查至关重要，而人工核查成本高昂。本文研究了引导式NLG中的自动化忠实度评估。我们开发了一个评分标准模板，并使用大语言模型（LLMs）在可量化的评分尺度上对生成内容进行打分。我们比较了主流大语言模型以及广泛采用的自然语言推理（NLI）模型在评分质量和敏感度方面的表现。此外，我们开发了合成不忠实数据的生成方法，以及量化幻觉百分比的启发式方法。我们在4个旅游领域行业数据集上的结果表明，GPT-4能够对源文本与生成内容之间是否事实一致给出准确的判断和解释。此外，我们发现使用合成数据微调NLI模型可以提升性能。最后，我们给出了部署此类系统在延迟和成本方面的见解。
英文摘要：
Hallucination has been a popular topic in natural language generation (NLG). In real-world applications, unfaithful content can result in poor data quality or loss of trust from end users. Thus, it is crucial to fact-check before adopting NLG for production usage, which can be expensive if done manually. In this paper, we investigate automated faithfulness evaluation in guided NLG. We developed a rubric template and used large language models (LLMs) to score the generation on quantifiable scales. We compared popular LLMs as well as widely adopted natural language inference (NLI) models in scoring quality and sensitivity. In addition, we developed methods for the generation of synthetic unfaithful data, as well as heuristics to quantify the percentage of hallucination. Our results on 4 travel-domain industry dataset show that GPT-4 can provide accurate judgement and explanation of whether a source and a generation are factually consistent. Furthermore, we found that tuning NLI models on synthetic data can improve performance. Lastly, we present insights on the latency and cost of deploying such a system.
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。
