Dataset · deepmind
AQUA-RAT: algebraic word problems with solutions
deepmind/aqua_rat
About 100k English algebraic word problems with step-by-step solution explanations.
What's inside
The dataset contains approximately 100,000 algebraic word problems. For each problem, the solution is explained step by step in natural language. The data was used to train a program generation model: the model learns to generate an explanation and a program solving the problem. Language is English.
What you can do with it
- Train an LLM to solve algebraic problems with explanations
- Evaluate question-answering and chain-of-thought on math
- Fine-tune a program generation model from a text problem
- Tasks
- Question answering
- Languages
- English
- Size
- 100K–1M rows
- Downloads this month
- 18.3K
- Likes
- 79
Models trained on it
- StrandsAgents/strands-decider-2B-qwen3.5-v1-2610Text classification 24.5K
- jaredpalmer/kev-27bText classification 3.4K
- mradermacher/kev-27b-GGUFMultiple choice 1.6K
- mradermacher/elf-1-12b-GGUF 737
- mradermacher/elf-1-31b-GGUF 710
- littlemex/prismyra-decision-35b-a3b-fp8Text classification 556
- blackmallangcow/nari-decision-4bEmbeddings 445
- littlemex/prismyra-decision-qwen3.6-35b-a3b-fp8-36lText classification 319
How to load it
from datasets import load_dataset
ds = load_dataset("deepmind/aqua_rat")