ClarusC64/clinical-electrolyte-instability-v0.1

该数据集用于评估模型是否能够检测由电解质失衡引起的电生理不稳定性。每个数据行代表三个时间点的简化临床轨迹，任务是确定电解质动态是否保持稳定或趋向于电生理不稳定性。数据集包含多个特征，如钾轨迹、钠轨迹、QT间期代理、肾功能代理、利尿剂负荷、纠正延迟和镁代理等，并包含一些干扰变量。预测目标是标签1表示电生理不稳定性风险，标签0表示稳定的电解质轨迹。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 electrophysiologic instability arising from electrolyte imbalance. Each row represents a simplified clinical trajectory across three time points. The task is to determine whether the electrolyte dynamics remain stable or move toward electrophysiologic instability. The dataset includes features such as potassium trajectory, sodium trajectory, QT interval proxy, renal function proxy, diuretic load, correction delay, and magnesium proxy, along with some decoy variables. The prediction target is label 1 for electrophysiologic instability risk and label 0 for stable electrolyte trajectory. The dataset is part of the Clarus Stability Reasoning Benchmark and is licensed under MIT.