Dataset · allenai
Science questions with a 17-million sentence corpus (QASC)
allenai/qasc
9,980 elementary and middle school science questions with 8 answer choices and a 17-million sentence corpus for reasoning training.
What's inside
The English-language QASC dataset evaluates sentence composition skills when answering elementary and middle school curriculum questions. It consists of 9,980 questions with 8 answer choices each (split into train, dev, and test sets) and is accompanied by a background text corpus of 17 million sentences.
What you can do with it
- Evaluating LLM multi-hop reasoning capabilities across sentences
- Training and testing RAG systems for retrieving required facts from a large corpus
- SFT fine-tuning of models on multiple-choice tasks
- Tasks
- Question answering, Multiple choice
- Languages
- English
- Size
- 1K–10K rows
- Downloads this month
- 22.5K
- Likes
- 24
Models trained on it
- jaredpalmer/kev-27bText classification 3.5K
- mradermacher/kev-27b-GGUFMultiple choice 1.7K
- mradermacher/elf-1-12b-GGUF 753
- mradermacher/elf-1-31b-GGUF 721
- littlemex/prismyra-decision-35b-a3b-fp8Text classification 553
- blackmallangcow/nari-decision-4bEmbeddings 441
- mradermacher/bobcat-1.3-GGUF 260
- sanghwa-na/bobcat-1.3Zero shot classification 239
How to load it
from datasets import load_dataset
ds = load_dataset("allenai/qasc")