JackHsieh/Qwen3-4B-Instruct-2507.rule-thoughtful-except-first.k-256.L-256.statml-arxiv

该数据集包含训练集和测试集，训练集有1,228,800个示例，测试集有19,200个示例。特征包括uuid（唯一标识符）、chunk_index（块索引）、chunk_start_index（块起始索引）、chunk_end_index（块结束索引）、g（未知整数值）、thought_token_ids（思维令牌ID列表）、thought_text（思维文本字符串）、truncated（截断标志）、finish_reason（完成原因字符串）、generated_at（生成时间戳）和generation_seconds（生成秒数）。数据集总大小约为3.06 GB，下载大小约为646.56 MB，配置为默认设置，数据文件路径分别对应训练和测试划分。This dataset includes train and test splits, with 1,228,800 examples in the train set and 19,200 examples in the test set. Features consist of uuid (unique identifier), chunk_index (chunk index), chunk_start_index (chunk start index), chunk_end_index (chunk end index), g (unknown integer value), thought_token_ids (list of thought token IDs), thought_text (thought text string), truncated (truncation flag), finish_reason (finish reason string), generated_at (generation timestamp), and generation_seconds (generation seconds). The total dataset size is approximately 3.06 GB, with a download size of about 646.56 MB, configured as default with data file paths corresponding to train and test splits.