ToolBridge

该数据集名为ToolBridge，旨在通过在精心挑选的数据条目上对大型语言模型（LLM）进行监督微调，使其能够有效利用外部工具。ToolBridge采用了一系列开放获取的数据集作为原始数据集池，并系统地识别适合插入外部工具API的数据条目。该数据集规模包含183,147次使用62个Python包的外部工具调用，其任务旨在辅助LLM使用外部工具进行数据处理、数值计算和事实检索的推理过程。

This dataset, named ToolBridge, aims to enable large language models (LLMs) to effectively utilize external tools via supervised fine-tuning on carefully selected data entries. ToolBridge adopts a collection of open-access datasets as its original dataset pool, and systematically identifies data entries suitable for integrating external tool APIs. This dataset comprises 183,147 external tool calls utilizing 62 Python packages, and its tasks are designed to assist LLMs in the reasoning process of using external tools for data processing, numerical computation, and factual retrieval.