Nemotron-RL-Agentic-Function-Calling-Pivot-v1-prompt-only
收藏资源简介:
Nemotron-RL-Agentic-Function-Calling-Pivot-v1-prompt-only 是一个专门从源数据集 nvidia/Nemotron-RL-Agentic-Function-Calling-Pivot-v1 中提取提示词(prompt)部分而创建的数据集。其核心是一个CSV文件(prompts.csv),每条记录包含提取出的用户提示(prompt)、分离的系统提示(system_prompt)以及当源数据定义了可用工具时的结构化工具描述(tools),嵌套值以JSON格式编码在CSV单元格内。数据集还包括两个元数据文件,用于总结提取统计信息(如行数统计、失败计数)和记录提取结果为空的源行索引。总共有9620条有效提取记录,无提取失败记录。该数据集适用于与提示词工程、基于大语言模型的函数调用或工具使用、以及强化学习智能体训练相关的任务,特别是作为仅包含提示信息的训练或评估数据。
Nemotron-RL-Agentic-Function-Calling-Pivot-v1-prompt-only is a dataset created specifically by extracting the prompt portions from the source dataset nvidia/Nemotron-RL-Agentic-Function-Calling-Pivot-v1. The core content is a CSV file (prompts.csv), where each record corresponds to a row in the source data and contains the extracted user prompt, separated system prompt, and structured tool descriptions (tools) when available tools are defined in the source data, with nested values encoded in JSON format within CSV cells. The dataset also includes two metadata files for summarizing extraction statistics (such as row counts, failure counts) and recording source row indices where extraction results were empty. There are a total of 9620 valid extraction records with no extraction failures. This dataset is suitable for tasks related to prompt engineering, large language model-based function calling or tool usage, and reinforcement learning agent training, particularly as training or evaluation data containing only prompt information.




