ClarusC64/clinical-autonomic-control-instability-v0.1

该数据集用于评估模型是否能检测由自主神经调节失败引起的不稳定性。每个数据行代表一个简化的自主神经控制场景，包含三个时间点。任务目标是判断自主神经调节是否保持稳定或趋向于调节不稳定。数据集的核心思想是通过反馈循环调节心率、血压和呼吸耦合来稳定生理状态。不稳定的情况包括心率快速上升、血压变异性增加、自主神经张力下降、压力反射缓冲减弱、生理压力增加或干预过晚。数据集的结构包括心率轨迹、血压变异性轨迹、自主神经张力代理、呼吸变异性代理、压力指数、压力反射代理和干预延迟等变量，并包含干扰变量如监测噪声和图表噪声。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Clarus稳定性推理基准的一部分，采用MIT许可证。

This dataset evaluates whether models can detect instability caused by autonomic regulation failure. Each row represents a simplified autonomic control scenario across three time points. The task is to determine whether autonomic regulation remains stable or is moving toward regulatory instability. The core stability idea is that the autonomic nervous system stabilizes physiology through feedback loops regulating heart rate, blood pressure, and respiratory coupling. Instability emerges when: heart rate rises rapidly, blood pressure variability increases, autonomic tone declines, baroreflex buffering weakens, physiological stress increases, or intervention occurs too late. The dataset includes variables such as heart rate trajectory, blood pressure variability trajectory, autonomic tone proxy, respiratory variability proxy, stress index, baroreflex proxy, and intervention delay, along with decoy variables like monitor_noise and chart_noise. Evaluation metrics include accuracy, precision, recall, f1, confusion matrix, and dataset integrity diagnostics. This dataset is part of the Clarus Stability Reasoning Benchmark and is licensed under MIT.