0

MGSM (Multilingual GSM8K)

Active

Hand-translated 250-problem subset of GSM8K in 10 languages - a multilingual grade-school math benchmark.

Open
Publisher
Google DeepMind
Capabilities
MathMultilingual
Domain
multilingual
Format
HF Dataset
Size
2500 tasks
License
MIT
Published
Oct 2022
Notable for
Benchmark for evaluating math and multilingual in the multilingual domain.
Canonical
github.com/google-research/url-nlp/tree/main/mgsm
Also on
huggingface.co/datasets/juletxara/mgsm

Cite

Notes

Only stored in your browser.

Related tools

2
View all

Implementations, trainers, datasets and scaffolds linked to this eval.

VF Openbench RL Env (Community)

Environment for single-turn tasks in OpenBench

Trains towardRL Env

Aya Dataset

Cohere

Cohere For AI's massively multilingual instruction dataset covering 65 languages, built by a 3,000-person open-science collaboration.

Training dataSFT DatasetInstruction FollowingMultilingual

FAQ

What is MGSM (Multilingual GSM8K)?
Hand-translated 250-problem subset of GSM8K in 10 languages - a multilingual grade-school math benchmark.
What capabilities does MGSM (Multilingual GSM8K) test?
MGSM (Multilingual GSM8K) evaluates math, multilingual.
How can a model improve its MGSM (Multilingual GSM8K) score?
Tools linked to MGSM (Multilingual GSM8K) on Sophon include VF Openbench RL Env (Community), Aya Dataset - RL environments, datasets, and scaffolds that target this eval.
What license is MGSM (Multilingual GSM8K) under?
MGSM (Multilingual GSM8K) is available under MIT.