AQuA-RAT: 100,000 Algebraic Word Problems with Multiple-Choice Answers and Step-by-Step Natural Language Rationales (DeepMind)

Provider
DeepMind (Ling et al.)
License
Apache-2.0 (source)
Size
8 files
Format
Parquet (in sampled files)
Stored at
Hugging Face (https://huggingface.co/datasets/deepmind/aqua_rat)
Access
Free and public. No account with the data provider is needed.
Category
Mathematics
Source
https://arxiv.org/abs/1705.04146

About this dataset

AQuA-RAT (Algebra Question Answering with Rationales) is a set of about 100,000 algebraic and arithmetic word problems, each with five multiple-choice options, the correct option and a rationale that explains the solution step by step in natural language mixed with equations. It was built by DeepMind to train models that generate an explanation and a program that solves the problem, and is now a common benchmark for math word problem solving, quantitative reasoning and chain-of-thought rationale generation. Splits: 97,467 training, 254 validation and 254 test problems. The repository has two configurations as Parquet files: raw/ (question, options, rationale, correct answer) and tokenized/ (the same fields pre-tokenized), about 52 MB in total. Load with the Hugging Face datasets library (load_dataset('deepmind/aqua_rat', 'raw')) or pandas. The data is published in the deepmind/AQuA GitHub repository under the Apache License 2.0; cite Ling et al., Program Induction by Rationale Generation: Learning to Solve and Explain Algebraic Word Problems, ACL 2017.

How to load this dataset in Python

from datasets import load_dataset
ds = load_dataset("deepmind/aqua_rat")
print(ds)

Data structure

File types in sampled files: Parquet

Files

Example file: raw/test-00000-of-00001.parquet

Analyze this dataset without downloading it

On Scigantic, AQuA-RAT: 100,000 Algebraic Word Problems with Multiple-Choice Answers and Step-by-Step Natural Language Rationales (DeepMind) opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.

This page is also available as Markdown and as JSON.

Related Mathematics datasets