Metric & Benchmark
GSM8K
Grade-school-math benchmark of 8.5k word problems with step-by-step solutions.
Definition
GSM8K is a 7.5k-train/1k-test set of grade-school maths word problems published by OpenAI. It became the canonical reasoning benchmark after chain-of-thought prompting unlocked dramatic improvements on it.
Common use cases
- Reasoning eval
- Chain-of-thought research