metr-evals/malt-transcripts-public

该数据集包含多个配置，每个配置下都有节点信息，节点信息中包含时间戳、消息内容、函数调用信息、元数据等。数据集还包含了标签、任务ID、模型名称、是否具有思维链等信息。数据集分为transcripts等多个分片，每个分片包含了大量的文本数据。

The dataset contains multiple configurations, each with node information that includes timestamps, message content, function call information, metadata, etc. The dataset also includes labels, task IDs, model names, and information on whether it has a chain of thought. The dataset is split into multiple segments such as transcripts, each containing a large amount of text data.