Skip to content

Dataset · HuggingFaceH4

Stack Exchange preferences for preference tuning

HuggingFaceH4/stack-exchange-preferences

Stack Overflow questions and answers with at least two responses for preference models and instruction tuning.

What's inside

Questions and answers from the Stack Overflow Data Dump. Questions are filtered to ensure each has at least two answers. English language, 10M to 100M records volume; the card also supports instruction fine-tuning and language model training.

What you can do with it

  • Training reward or preference models
  • Fine-tuning LLMs on instructions and Stack Overflow answers
  • Evaluating quality of technical question responses
Tasks
Question answering
Languages
English
Size
10M–100M rows
Downloads this month
13.7K
Likes
140

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("HuggingFaceH4/stack-exchange-preferences")