axolotl-ai-co/numina-cot-logprobs-pipeline-samples

numina-cot-logprobs-pipeline-samples数据集是一个合成数据集，包含用户和助手之间的对话。每个样本包括对话内容、角色标识以及对应的日志概率信息。数据集结构包括对话内容（content）、角色（role）、prompt、distilabel_metadata等字段。distilabel_metadata中包含原始输入和输出的日志概率，以及统计信息。该数据集适用于自然语言处理任务，如对话系统、语言模型训练等。

The numina-cot-logprobs-pipeline-samples dataset is a synthetic dataset containing conversations between users and assistants. Each sample includes the conversation content, role identification, and corresponding log probability information. The dataset structure includes fields such as content, role, prompt, distilabel_metadata. Distilabel_metadata contains raw input and output log probabilities, as well as statistical information. This dataset is suitable for natural language processing tasks such as dialogue systems and language model training.