Dataset · allenai
SciQ science multiple-choice questions
allenai/sciq
13,679 crowd-sourced physics, chemistry, and biology exam questions with 4 answer choices and explanations.
What's inside
The dataset contains 13,679 science questions in a multiple-choice format, each with four answer choices. Most questions include a paragraph explaining the correct answer. The crowd-sourced questions cover physics, chemistry, biology, and other sciences in English.
What you can do with it
- Fine-tune a QA model or LLM to answer science questions
- Evaluate model reasoning and accuracy on SciQ
- Collect SFT examples with answer explanations
- Tasks
- Question answering
- Languages
- English
- Size
- 10K–100K rows
- Downloads this month
- 120K
- Likes
- 170
Models trained on it
- trjxter/Gwimi-4-12B-IT-GGUFText generation 1.1K
- IvmeLabs/Ivme-Conversate-N-v1-BaseText generation 687
- ReckonerLabs/Pascaline-3MText generation 645
- mradermacher/MiniCPM4-8B-PaperProf-i1-GGUF 571
- littlemex/prismyra-decision-qwen3.6-35b-a3b-fp8-36lText classification 321
- littlemex/prismyra-decision-qwen3.6-35b-a3b-nvfp4-36lText classification 245
- mradermacher/MiniCPM4-8B-PaperProf-GGUF 197
- OpenSML/Solstice-1-175MText generation 177
How to load it
from datasets import load_dataset
ds = load_dataset("allenai/sciq")