Skip to content

Dataset · nvidia

Nemotron-Math-v2: mathematical reasoning

nvidia/Nemotron-Math-v2

347K math problems and 7M model-generated reasoning trajectories in English.

What's inside

The dataset contains about 347K high-quality math problems and 7M model-generated reasoning trajectories. These are English text data for long-context distillation of mathematical reasoning from multi-mode supervision. It accompanies the Nemotron-Math paper; code and documentation are available in NeMo-Skills.

What you can do with it

  • Train LLMs in mathematical reasoning
  • Distill long-context reasoning
  • Evaluate models on math tasks
Tasks
Text generation
Languages
English
Size
1M–10M rows
Downloads this month
17.2K
Likes
204

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("nvidia/Nemotron-Math-v2")