Skip to content

Dataset · dvilasuero

Compact dataset for natural science reasoning

dvilasuero/natural-science-reasoning

A small open-source set of English prompts and answers for reproducible experiments with reasoning models.

What's inside

The repository contains English text prompts and pipeline configuration (fewer than 1,000 records) generated via Hugging Face Inference Endpoints using models like DeepSeek V3, DeepSeek-R1, and Llama 70B-Instruct. The dataset is intentionally small and focused on natural sciences in addition to classical math and coding tasks.

What you can do with it

  • Testing and reproducing chain-of-thought reasoning generation pipelines
  • Fine-tuning models on natural science reasoning
Tasks
Text generation
Languages
English
Size
up to 1K rows
Downloads this month
1.2K
Likes
40

How to load it

from datasets import load_dataset

ds = load_dataset("dvilasuero/natural-science-reasoning")