Datasets:

TTimur
/

kyrgyzMMLU

configs:
  - config_name: kyrgyz_mmlu_all
    data_files:
      - split: validation
        path: kyrgyz_mmlu_all/validation.parquet
      - split: test
        path: kyrgyz_mmlu_all/test.parquet
  - config_name: kyrgyz_mmlu_biology
    data_files:
      - split: validation
        path: kyrgyz_mmlu_biology/validation.parquet
      - split: test
        path: kyrgyz_mmlu_biology/test.parquet
  - config_name: kyrgyz_mmlu_chemistry
    data_files:
      - split: validation
        path: kyrgyz_mmlu_chemistry/validation.parquet
      - split: test
        path: kyrgyz_mmlu_chemistry/test.parquet
  - config_name: kyrgyz_mmlu_geography
    data_files:
      - split: validation
        path: kyrgyz_mmlu_geography/validation.parquet
      - split: test
        path: kyrgyz_mmlu_geography/test.parquet
  - config_name: kyrgyz_mmlu_history
    data_files:
      - split: validation
        path: kyrgyz_mmlu_history/validation.parquet
      - split: test
        path: kyrgyz_mmlu_history/test.parquet
  - config_name: kyrgyz_mmlu_lang
    data_files:
      - split: validation
        path: kyrgyz_mmlu_lang/validation.parquet
      - split: test
        path: kyrgyz_mmlu_lang/test.parquet
  - config_name: kyrgyz_mmlu_literature
    data_files:
      - split: validation
        path: kyrgyz_mmlu_literature/validation.parquet
      - split: test
        path: kyrgyz_mmlu_literature/test.parquet
  - config_name: kyrgyz_mmlu_math
    data_files:
      - split: validation
        path: kyrgyz_mmlu_math/validation.parquet
      - split: test
        path: kyrgyz_mmlu_math/test.parquet
  - config_name: kyrgyz_mmlu_medicine
    data_files:
      - split: validation
        path: kyrgyz_mmlu_medicine/validation.parquet
      - split: test
        path: kyrgyz_mmlu_medicine/test.parquet
  - config_name: kyrgyz_mmlu_physics
    data_files:
      - split: validation
        path: kyrgyz_mmlu_physics/validation.parquet
      - split: test
        path: kyrgyz_mmlu_physics/test.parquet
tags:
  - kyrgyz
  - llm-benchmark
  - evaluation
  - mmlu
  - multiple-choice
  - kyrgyzllm-bench
language:
  - ky

KyrgyzMMLU

KyrgyzMMLU is a native, multiple-choice benchmark for evaluating Large Language Models in the Kyrgyz language. It contains 7,977 questions written by Kyrgyz curriculum experts, covering 9 subjects relevant to the national curriculum and public knowledge.

🏔️ Part of the KyrgyzLLM-Bench

This dataset is a core component of the KyrgyzLLM-Bench, a comprehensive suite for evaluating LLMs in Kyrgyz.

Main Paper: Bridging the Gap in Less-Resourced Languages: Building a Benchmark for Kyrgyz Language Models
Hugging Face Hub: https://huggingface.co/TTimur
GitHub Project: [https://github.com/golden-ratio/kyrgyzLLM_bench]

📋 Dataset Description

This benchmark is designed to evaluate deep knowledge across a wide range of domains.

Subjects Covered:

Math: 1,169
Physics: 1,228
Geography: 640
Biology: 1,550
Kyrgyz Language: 360
Kyrgyz Literature: 1,169
Kyrgyz History: 440
Medicine: 216
Chemistry: 1,205

📜 Citation

If you find this dataset useful in your research, please cite the main project paper:

@article{KyrgyzLLM-Bench,
  title={Bridging the Gap in Less-Resourced Languages: Building a Benchmark for Kyrgyz Language Models},
  author={Timur Turatali, Aida Turdubaeva, Islam Zhenishbekov, Zhoomart Suranbaev, Anton Alekseev, Rustem Izmailov},
  year={2025},
  url={[https://huggingface.co/datasets/TTimur/kyrgyzMMLU]}
}