ClarusC64/clinical-intervention-delay-failure-v0.1

该数据集用于评估模型是否能检测由临床干预延迟导致的失败情况。每个场景描述了一个简短的生理轨迹以及治疗时间信号。临床恶化通常不是因为某个参数极端，而是因为治疗相对于系统轨迹到达得太晚。数据集要求模型综合考虑压力轨迹、代谢应激轨迹、心率趋势、治疗延迟和液体反应能力等因素。预测目标是判断是否因干预延迟导致不稳定（label=1）或轨迹稳定或被挽救（label=0）。数据集结构包括MAP轨迹、乳酸轨迹、心率轨迹、治疗延迟变量、液体反应指标和肾脏应激标记等，同时包含一些干扰变量。评估指标包括准确率、精确率、召回率、F1分数、混淆矩阵和数据集完整性诊断。该数据集是ClarusC64稳定性推理基准系列的一部分，支持跨域推理和稳定性检测的研究。

This dataset evaluates whether models can detect failure driven by delayed clinical intervention. Each scenario describes a short physiological trajectory along with treatment timing signals. Clinical deterioration often occurs not because a parameter is extreme, but because treatment arrives too late relative to the system trajectory. The dataset requires reasoning across pressure trajectory, metabolic stress trajectory, heart rate trend, treatment delay, and fluid response capacity. The prediction target is label = 1 for instability due to delayed intervention and label = 0 for trajectory stabilized or rescued. The row structure includes MAP trajectory, lactate trajectory, heart rate trajectory, treatment delay variables, fluid response indicator, and renal stress marker, along with decoy variables. Evaluation metrics include accuracy, precision, recall, f1, confusion matrix, and dataset integrity diagnostics. This dataset is part of the ClarusC64 stability-reasoning benchmark family and supports research into cross-domain reasoning and stability detection.