lowry02/hs-layer-5

该数据集是一个用于自然语言处理（NLP）任务的数据集，包含训练、验证和测试三个分割。每个样本具有多个特征，包括数据集样本ID、分割类型、层索引、token位置、token ID、token标签、逻辑标签以及隐藏状态（以float16列表形式存储）。隐藏状态可能来自语言模型的中间层输出，适用于分析模型内部表示或进行下游任务如分类或序列标注。数据集总大小约为1.78 GB，下载大小约为1.77 GB，其中训练集包含351,242个示例，验证集和测试集各包含约39,890和39,892个示例。

This dataset is designed for natural language processing (NLP) tasks, comprising train, validation, and test splits. Each sample includes multiple features such as dataset sample ID, split type, layer index, token position, token ID, token label, logic label, and hidden state (stored as a list of float16 values). The hidden states likely represent intermediate outputs from language model layers, suitable for analyzing model internal representations or performing downstream tasks like classification or sequence labeling. The total dataset size is approximately 1.78 GB with a download size of about 1.77 GB, containing 351,242 examples in the train split, and around 39,890 and 39,892 examples in the validation and test splits, respectively.