Skip to content

Dataset · lockon

Synthetic dialogues for training LLMs to use APIs

lockon/ToolACE

A text dialogue dataset in English and Chinese using 26,500 different APIs for agent training.

What's inside

Created via a multi-agent data synthesis pipeline based on a pool of 26,507 diverse APIs, the dataset contains multi-turn dialogues in English and Chinese with formalized reasoning processes, verified through a two-level rule- and model-based accuracy check. The repository contains 10k to 100k records.

What you can do with it

  • SFT fine-tuning of language models for function calling
  • Creating AI agents capable of planning action chains across complex APIs
Tasks
Text generation
Languages
English, Chinese
Size
10K–100K rows
Downloads this month
31.4K
Likes
1

How to load it

from datasets import load_dataset

ds = load_dataset("lockon/ToolACE")