Natural Hallucination Dataset - ACI-bench Clinical Note Hallucination Annotations
收藏资源简介:
该存储库包含来自ACI-bench数据集的专家标注的幻觉标签,用于评估医学文本摘要中的幻觉检测。自然幻觉(NH)数据集包含临床摘要中幻觉的专家标注,重点关注ACI-bench临床对话集中的SOAP笔记。
This repository contains expert-annotated delusion labels from the ACI-bench dataset, designed for evaluating delusion detection in medical text summarization. The Natural Delusion (NH) dataset includes expert annotations of delusions in clinical summaries, with a focus on the SOAP notes within the ACI-bench clinical dialogue corpus.
Natural Hallucination Dataset - ACI-bench Clinical Note Hallucination Annotations 数据集概述
数据集简介
- 数据集包含专家标注的临床摘要中的幻觉标签,专注于ACI-bench临床对话集合中的SOAP笔记。
- 用于评估医学文本摘要中的幻觉检测。
标注类别与数量
- No Error: 12,365
- Hallucination: 106
- Inference: 87
- Misunderstanding: 72
错误严重程度分布
- Low Severity: 138
- High Severity: 87
- Not Medically Relevant (NMR): 40
高严重性错误类别
- Diagnosis
- Exam Findings
- Lab Testing and Imaging
- Medical History
- Symptoms
- Treatment Plan
数据集格式
- 原始ACI-bench对话记录
- 专家标注的事实错误类别
- 每个错误的严重性标签
- 每个主题的聚合错误分数
用途
- 评估幻觉检测方法
- 分析临床摘要中不同类型的事实错误
- 研究医学文本生成中的高严重性与低严重性错误
引用
@inproceedings{BN2024fact, title={Fact-Controlled Diagnosis of Hallucinations in Medical Text Summarization}, author={BN, Suhas and Shing, Han-Chin and Xu, Lei and Strong, Mitch and Burnsky, Jon and Ofor, Jessica and Mason, Jordan R and Chen, Susan and Srinivasan, Sundararajan and Shivade, Chaitanya and Moriarty, Jack and Cohen, Joseph Paul}, booktitle={Interspeech}, year={2025}, organization={ISCA} }
注意事项
- 该发布仅包含ACI Bench摘要的专家标注,由于许可问题,LLM输出未公开。




