Tool Interaction Dataset
收藏arXiv2025-09-30 收录
下载链接：
https://github.com/public-apis/public-apis
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是通过在模拟环境中，利用公共API，让LLM智能体与各种工具集互动生成的，旨在评估FICAL算法的性能。互动过程的追踪由一个评判LLM（DeepSeek-v2）进行评估，以确定工具使用的正确性。该数据集的规模涉及5个客户端，每个客户端配备5套工具集。任务目标是评估LLM智能体在使用工具时的正确性和有效性。
This dataset was generated by having LLM Agents interact with various tool sets via public APIs in a simulated environment, with the primary objective of evaluating the performance of the FICAL algorithm. The interaction process is tracked and evaluated by a judge LLM (DeepSeek-v2) to determine the correctness of tool usage. This dataset covers 5 clients, each equipped with 5 tool sets. The task goal is to assess the correctness and effectiveness of LLM Agents when utilizing tools.
