datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
mgsm-gold
MGSM Gold - Multilingual Grade School Math
This dataset contains the MGSM (Multilingual Grade School Math) benchmark - 250 math word problems translated into 10 languages.
Attribution
This dataset is derived from juletxara/mgsm
Original source: google-research/url-nlp/mgsm
Usage
from datasets import load_dataset
# Load German test set
dataset = load_dataset("alibashir/mgsm-gold", "de")
print(dataset["test"][0])
Languages
Code
Language… See the full description on the dataset page: https://huggingface.co/datasets/alibashir/mgsm-gold.mgsm_ca
Dataset Card for mgsm_ca
mgsm_ca is a question answering dataset in Catalan that has been professionally translated from the MGSM dataset in English.
Dataset Details
Dataset Description
mgsm_ca (Multilingual Grade School Math - Catalan) is designed to evaluate multi-step mathematical reasoning using grade school math word problems. It includes 8 instances in the train split and another 250
instances in the test split. Each instance contains a math problem… See the full description on the dataset page: https://huggingface.co/datasets/projecte-aina/mgsm_ca.mgsm_gl
Dataset Card for mgsm_gl
mgsm_gl is a question answering dataset in Galician translated from the MGSM dataset in English.
Dataset Details
Dataset Description
This dataset is the Galician version of the MGSM (Multilingual Grade School Math) dataset. It serves as a benchmark of grade-school math problems as proposed in the paper Language models are multilingual chain-of-thought reasoners. It includes 8 instances in the train split and another 250 instances in the… See the full description on the dataset page: https://huggingface.co/datasets/proxectonos/mgsm_gl.MGSM-Symbolic
MGSM-Symbolic
MGSM-Symbolic is a multilingual symbolic variant of the Multilingual Grade School Math Benchmark (MGSM).It contains mathematically structured word problems across multiple languages, paired with numerical solutions.
The dataset is designed to support research in:
Multilingual reasoning
Cross-lingual generalisation
Symbolic numerical problem solving
Evaluation of reasoning consistency across languages
Each language contains the same set of problems translated and… See the full description on the dataset page: https://huggingface.co/datasets/lrana/MGSM-Symbolic.mgsm-rev2
MGSM-Rev2
MGSM-Rev2 is a corrected version of the MGSM benchmark, which evaluates multilingual mathematical reasoning on grade-school word problems across 10 languages.
Please refer to the original repository for details.
filtered_mgsm_with_ids
Filtered MGSM with IDs
This dataset is a filtered subset of [juletxara/mgsm] with an added integer id per language.
English questions overlapping with a PolyMath English set were removed, and the same ids were excluded from all languages to avoid cross-language overlaps.
Source: juletxara/mgsm (test split)
Filtering date: 2025-09-15
Splits: each language is exposed as a split (en, bn, de, es, ja, sw, te, th).
Fields
id (int): per-language index assigned before… See the full description on the dataset page: https://huggingface.co/datasets/deokhk/filtered_mgsm_with_ids.mgsm
MGSM – Urdu
Machine-translated Urdu subset of the Multilingual Grade School Math (MGSM) benchmark. This dataset was translated using Google Machine Translate.
Dataset Description
MGSM is a multilingual extension of the GSM8K math reasoning benchmark. This repository contains the Urdu translation, providing both a train split (8 few-shot exemplars) and a test split (250 problems).
Each example consists of a grade-school math word problem in Urdu along with its numerical… See the full description on the dataset page: https://huggingface.co/datasets/Owos/mgsm.mgsm-rev2-with-train
MGSM-Rev2 (with CoT train split)
This dataset regroups, per language, the corrected test set of
lightonai/mgsm-rev2 together with the
chain-of-thought few-shot train exemplars of
juletxara/mgsm, so that both are available from a
single dataset (the original MGSM-Rev2 ships only the test set, without a train split).
Splits
train: the 8 standard MGSM chain-of-thought exemplars per language (from juletxara/mgsm),
usable as few-shot demonstrations.
test: the 250… See the full description on the dataset page: https://huggingface.co/datasets/OpenLLM-France/mgsm-rev2-with-train.