利用低级符号能力实现幻觉检测中的无监督接地
Leveraging Low-Level Symbolic Competences for Unsupervised Grounding in Hallucination Detection
浏览论文内容
中文总结 AI 辅助
本研究利用LLM的SQL低级符号能力构建参考文档数据库，实现无监督幻觉检测，在RAGTruth、DiaHalu数据集上效果优于直接预测，可与SOTA方法竞争且无需领域微调。
中文摘要 AI 辅助
幻觉是指语言模型生成事实上不正确或无来源支持的输出，这对提示式和微调后的语言模型都是重大挑战。由于大型语言模型（LLM）的推理过程不透明，难以解释输出为何不准确，导致幻觉检测困难。本研究探究LLM是否可利用SQL这类低级符号能力，在高级任务中实现无监督幻觉检测。为此，我们让LLM从参考文档构建SQL数据库，该数据库用于在基于数据库接地的幻觉检测流程中，对参考内容和采样响应进行推理，提供神经符号检查。在RAGTruth和DiaHalu幻觉检测数据集上，我们发现该方法优于直接预测，可与最先进的幻觉检测方法竞争，且无需特定领域微调，仅依赖LLM已具备的低级通用能力，这值得进一步研究神经符号方法中LLM的低级能力。
英文摘要
Hallucination-where a language model generates outputs that are factually incorrect or unsupported by the source-is a major challenge for both prompted and fine-tuned language models. Detecting hallucinations is difficult due to the opaque reasoning processes of LLMs, which often provide little insight into why a model's output may be inaccurate. In this work, we investigate whether an LLM can use an alternative, low level, symbolic competence such as SQL for unsupervised hallucination detection in some high level task. For this, we make an LLM build an SQL database from reference documents. This SQL database is then used for reasoning over the reference and the sampled response in a hallucination detection pipeline that is grounded in the database, thereby providing a neurosymbolic checkup. On RAGTruth and DiaHalu hallucination detection datasets, we find that our approach improves on direct prediction and competes with state-of-the-art hallucination detection methods, while not requiring domain-specific fine-tuning. Instead it relies on a low-level general competence already present in LLMs. This warrants further investigation of low-level LLM competences in neurosymbolic approaches.
发表机构
- Heinrich Heine University Düsseldorf（杜塞尔多夫海因里希·海涅大学）
机构由 AI 辅助整理，请以论文原文为准。
