MBZUAI/KazMMLU

MukhammedTogmanov/jana

This dataset contains multiple subsets for different subjects and domains, structured for few-shot learning and evaluation.

Dataset Structure

The dataset is organized into subsets, each corresponding to a specific subject and level. Each subset has two splits:

• —`dev`: Few-shot examples (3 rows per subset).
• —`test`: The remaining rows for evaluation.

Subsets

The dataset contains the following subsets:

• —Accounting (University)
• —Biology (High School in kaz)
• —Biology (High School in rus)
• —Chemistry (High School in kaz)
• —Chemistry (High School in rus)
• —... (list other subsets)

Splits

Features

The dataset includes the following columns:

• —`Number`: Question ID.
• —`Source`: URL or source of the question.
• —`Group`: The group to which the subject belongs (e.g., Social Science, Science).
• —`Subject`: The subject of the question (e.g., Biology, Accounting).
• —`Level`: Academic level (e.g., High School, University).
• —`Question`: The text of the question.
• —`Option A`, `Option B`, `Option C`, `Option D`, `Option E`: The answer choices.
• —`Answer Key`: The correct answer.
• —`Subset`: The name of the subset (e.g., Biology (High School in kaz)).
• —`Split`: The data split (dev or test).
• —`is_few_shot`: Indicates whether the row is part of the few-shot examples (1 for dev, 0 for test).

Usage

You can load this dataset with the following code:

from datasets import load_dataset

dataset = load_dataset("MukhammedTogmanov/jana")