Skip to content

Dataset · TIGER-Lab

Web instructions from forums

TIGER-Lab/WebInstructSub

High-quality instruction subset from web forums for LLM training.

What's inside

An English text dataset ranging from 1 million to 10 million examples. This is part of the data from MAmmoTH2, collected mainly from forums like StackExchange. The subset is positioned as high-quality instructions to improve LLM quality.

What you can do with it

  • SFT / instruction fine-tuning of LLMs
  • Training question-answering models
  • LoRA for reasoning tasks
Tasks
Question answering
Languages
English
Size
1M–10M rows
Downloads this month
8.9K
Likes
173

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("TIGER-Lab/WebInstructSub")