Skip to content

Dataset · allenai

650k Human-ChatGPT Dialogues

allenai/WildChat

Multilingual real user dialogues with ChatGPT for chat model training and analysis.

What's inside

WildChat is a collection of 650k dialogues between humans and ChatGPT. The data is multilingual, suitable for text-generation and question-answering tasks. The card indicates the ODC-BY license and a link to a newer version with 4.8 million dialogues and demographic information.

What you can do with it

  • Train a chat model on real dialogues
  • Test a QA system
  • Study ChatGPT user behavior
Tasks
Text generation, Question answering
Size
100K–1M rows
Downloads this month
6.1K
Likes
242

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("allenai/WildChat")