arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2501.13951cs.CLcs.AI
收藏

一种用于长上下文心理健康评估的分层多专家框架

A Layered Multi-Expert Framework for Long-Context Mental Health Assessments

Texas A\&M University-Corpus Christi(德克萨斯A&M大学-科珀斯克里斯蒂);Delft University of Technology(代尔夫特理工大学)

Jinwen Tang, Qiming Guo, Wenbo Sun, Yi Shang

AI总结:

本文提出分层多专家框架SMMR,通过多LLM与专用小模型分阶段处理长上下文心理健康评估,在DAIC-WOZ等数据上显著提升准确率并减少幻觉。

摘要:

长文本心理健康评估对大语言模型(LLMs)构成独特挑战,这些模型在处理扩展的、领域特定的上下文时往往表现出幻觉或不一致的推理。我们提出了堆叠式多模型推理(SMMR),这是一个分层框架,将多个LLMs和专门的较小模型作为平等的“专家”加以利用。早期层隔离出短小、离散的子任务,而后期层通过更先进的长上下文模型整合并细化这些部分输出。我们在DAIC-WOZ抑郁筛查数据集和48个经过整理的精神病诊断案例研究上评估了SMMR,结果表明在准确率、F1分数和PHQ-8误差降低方面,相较于单模型基线有一致的改进。通过利用多样化的“第二意见”,SMMR减轻了幻觉,捕捉了细微的临床特征,并增强了高风险心理健康评估中的可靠性。我们的发现强调了多专家框架对于更值得信赖的AI驱动筛查的价值。

英文摘要:

Long-form mental health assessments pose unique challenges for large language models (LLMs), which often exhibit hallucinations or inconsistent reasoning when handling extended, domain-specific contexts. We introduce Stacked Multi-Model Reasoning (SMMR), a layered framework that leverages multiple LLMs and specialized smaller models as coequal 'experts'. Early layers isolate short, discrete subtasks, while later layers integrate and refine these partial outputs through more advanced long-context models. We evaluate SMMR on the DAIC-WOZ depression-screening dataset and 48 curated case studies with psychiatric diagnoses, demonstrating consistent improvements over single-model baselines in terms of accuracy, F1-score, and PHQ-8 error reduction. By harnessing diverse 'second opinions', SMMR mitigates hallucinations, captures subtle clinical nuances, and enhances reliability in high-stakes mental health assessments. Our findings underscore the value of multi-expert frameworks for more trustworthy AI-driven screening.

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。

↑