CogInstruct
收藏资源简介:
CogInstruct是一个用于可解释心理压力检测的指令调优数据集,由牛津大学工程科学系等机构开发。该数据集基于认知评估理论,通过三阶段的自我反思标注流程生成,旨在帮助大型语言模型(LLMs)生成逐步推理的压力检测解释。数据集的内容包括从刺激到评估、反应再到压力状态的认知链,适用于提升压力检测模型的可解释性和性能。
CogInstruct is an instruction-tuning dataset for explainable psychological stress detection, developed by the Department of Engineering Science of the University of Oxford and other institutions. Grounded in cognitive appraisal theory, this dataset is generated via a three-stage self-reflective annotation workflow, aiming to assist large language models (LLMs) in generating step-by-step reasoning explanations for stress detection. The dataset includes a cognitive chain spanning from stimulus, through appraisal and reaction, to stress state, and is designed to enhance the interpretability and performance of stress detection models.

- 1Cognition Chain for Explainable Psychological Stress Detection on Social Media牛津大学工程科学系 · 2024年



