Skip to content

Dataset · nvidia

Multi-domain RL reasoning dataset

nvidia/Nemotron-CrossThink

10-100M examples for RL training of LLMs on general and mathematical reasoning across domains.

What's inside

Nemotron-CrossThink is a multi-domain reinforcement learning dataset designed to improve general and mathematical reasoning in large language models. The data is textual, in English, with a size ranging from 10 to 100 million examples. The dataset has a paper and blog post; authors are listed in the card.

What you can do with it

  • RL fine-tuning of LLMs for reasoning
  • Improving mathematical and general reasoning skills
  • Evaluating self-learning approaches beyond mathematics
Tasks
Question answering, Text generation
Languages
English
Size
10M–100M rows
Downloads this month
1.3K
Likes
114

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("nvidia/Nemotron-CrossThink")