GSM8K Math Reasoning
Dataset
GSM8K (Grade School Math 8K) is a high-quality benchmark dataset of elementary school math problems created by OpenAI, expanded to 17.6K problems with detailed step-by-step solutions, and has become the standard benchmark for evaluating the mathematical reasoning abilities of large language models. Each problem requires 2-8 basic mathematical operations to solve.
Dataset Highlights
The gold standard benchmark dataset for evaluating the mathematical reasoning abilities of large language models
Elementary Math Problems
The questions cover basic mathematical operations at the elementary level, including addition, subtraction, multiplication, division, fractions, percentages, etc., requiring no advanced mathematical knowledge, focusing on testing reasoning rather than computational complexity.
Step-by-Step Solutions
Each question comes with a detailed step-by-step solution process, highlighting intermediate calculation steps and the final answer, providing high-quality annotated data for Chain-of-Thought training.
Language Diversity
The questions use natural language to describe a rich variety of life scenarios, covering themes such as shopping, speed, age, etc., with flexible language expression to avoid formulaic patterns, making them closer to real-world problems.
Multi-Step Reasoning
Each question requires 2-8 reasoning steps to answer, demanding the model to have coherent logical reasoning and intermediate state tracking abilities, effectively distinguishing the reasoning levels of different models.
Natural Language Questions
All questions are presented in natural language text form, without formulas or symbolic expressions, testing the model's ability to understand mathematical relationships from text and extract key numerical values.
Standard Benchmark
Widely used by mainstream large language models such as GPT-4, Claude, and Gemini for evaluating mathematical reasoning abilities, it is one of the most cited mathematical benchmarks in academic papers and technical reports.
Applicable Scenarios
From model evaluation to educational research and development, fully supports mathematical reasoning research
Mathematical Reasoning Assessment
Assess the multi-step mathematical reasoning ability of large language models, quantifying the model's accuracy and reasoning quality on basic math problems
Chain of Thought Training
Train Chain-of-Thought reasoning using step-by-step answer data to enhance the model's ability to break down complex problems
LLM Benchmark Testing
Serve as a standardized benchmark to compare the mathematical reasoning performance of different models, tracking the trend of capability changes between model iterations
Educational AI Development
Build intelligent math tutoring systems and automated problem-solving tools, providing training and evaluation data for AI applications in the K-12 education sector
Data Preview
The following are typical questions and step-by-step answer examples from the GSM8K dataset
{
"question": "Natalia sold clips to 48 of her friends in April, and then she sold half as many clips in May. How many clips did Natalia sell altogether in April and May?",
"answer": "Natalia sold 48/2 = 24 clips in May.\nNatalia sold 48 + 24 = 72 clips altogether in April and May.\n#### 72"
}
{
"question": "Betty is saving money for a new wallet which costs $100. Betty has only half of the money she needs. Her parents decided to give her $15 for that purpose, and her grandparents twice as much as her parents. How much more money does Betty need to buy the wallet?",
"answer": "In the beginning, Betty has only 100 / 2 = $50.\nBetty's grandparents gave her 15 * 2 = $30.\nThis means, Betty needs 100 - 50 - 30 - 15 = $5 more.\n#### 5"
}
3 Steps to Get Started Quickly
From browsing to usage, you can start your mathematical reasoning research in just a few minutes
Browse the Dataset
View the details of the GSM8K dataset on the Ace Data Cloud platform, and understand metadata such as question format, answer structure, and licensing agreements.
Get the Data
Obtain a complete set of 17.6K math problems and step-by-step solutions via API, with standardized data format, ready to use.
Evaluation and Training
Use the dataset to evaluate the model's mathematical reasoning ability, or as a training data source for Chain-of-Thought fine-tuning.
Start Exploring the GSM8K Mathematical Reasoning Data
A standard mathematical reasoning benchmark produced by OpenAI, available under MIT open license, get it now. Whether you are evaluating large model capabilities or training reasoning models, GSM8K is an indispensable dataset.