BLWX/ToolBridge

ToolBridge是一个开源数据集，旨在帮助大型语言模型（LLM）学习如何有效地使用外部工具。数据集由三种类型的条目组成：原始对话数据，包含Python代码插入的对话数据，以及包含代码插入和执行结果的对话数据。通过在监督微调（SFT）过程中使用这些经过精心挑选的数据条目，LLM可以学会在适当的上下文中调用外部工具，从而提高其在数值计算、数据处理和事实检索等关键功能上的预测准确性。

ToolBridge is an open-source dataset designed to assist large language models (LLMs) in learning how to effectively use external tools. The dataset consists of three types of entries: original dialogue data, dialogue data with Python code insertions, and dialogue data with code insertions and execution results. By utilizing these carefully selected data entries in the supervised fine-tuning (SFT) process, LLMs can learn to invoke external tools in appropriate contexts, thereby enhancing their predictive accuracy on essential functions such as numerical computation, data processing, and factual retrieval.