Skip to content

Dataset · TIGER-Lab

Verified instructions for reasoning

TIGER-Lab/WebInstruct-verified

A set of English text questions and instructions for training LLM reasoning.

What's inside

WebInstruct-verified is a text dataset in English ranging from 100k to 1M examples. The card links it to General-Reasoner: training on diverse verifiable questions with a model-based verifier. The data is designed to develop LLM reasoning across different domains.

What you can do with it

  • Train LLMs to reason on questions with verifiable answers
  • Test a verifier model for reasoning selection
  • Fine-tune a QA model via SFT
Tasks
Question answering
Languages
English
Size
100K–1M rows
Downloads this month
12.5K
Likes
69

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("TIGER-Lab/WebInstruct-verified")