Skip to content

Dataset · TNSA

Synthetic reasoning and instruction corpus (1B–10B)

TNSA/PT-HF500B

A large English FinePhrase synthetic dataset for training LLMs in logic and instruction following.

What's inside

The dataset contains between 1 billion and 10 billion English examples in text and tabular formats. It consists of educational web data reformatted into a structured synthetic dataset with instructions; it was used to pre-train the NGen-3 and NGen-4 model families.

What you can do with it

  • Pre-training large language models
  • Improving logical reasoning skills in LLMs
  • Training models to accurately follow complex instructions
Tasks
Text generation
Languages
English
Size
1B–10B rows
Downloads this month
25.7K
Likes
1

How to load it

from datasets import load_dataset

ds = load_dataset("TNSA/PT-HF500B")