Datasets:

f1rdavs
/

TajLiB

dataset_info:
  features:
    - name: id
      dtype: string
    - name: question
      dtype: string
    - name: option_a
      dtype: string
    - name: option_b
      dtype: string
    - name: option_c
      dtype: string
    - name: option_d
      dtype: string
    - name: answer
      dtype: string
  splits:
    - name: grammar
      num_bytes: 330269
      num_examples: 919
    - name: law
      num_bytes: 135655
      num_examples: 336
    - name: literature
      num_bytes: 178133
      num_examples: 320
  download_size: 294150
  dataset_size: 644057
configs:
  - config_name: default
    data_files:
      - split: grammar
        path: data/grammar-*
      - split: law
        path: data/law-*
      - split: literature
        path: data/literature-*
license: mit
task_categories:
  - question-answering
  - multiple-choice
tags:
  - tajik
  - linguistics
  - ntc
  - education
pretty_name: TajLiB
size_categories:
  - n<2K
language:
  - tg

Tajik Linguistic Benchmark (TajLiB)

Dataset Description
Dataset Structure
- Data Instances
- Data Fields
Dataset Creation
- Source Data
- Curation Process
Considerations for Using the Data
Citation Information

Dataset Description

TajLiB (Tajik Linguistic Benchmark) is a comprehensive multiple-choice question benchmark designed to evaluate the linguistic understanding of Large Language Models (LLMs) in the Tajik language.

It assesses the model's knowledge of Tajik phonetics, grammar, lexicology, and orthography, as well as its ability to understand and reason about domains related to Tajik language usage, including legal texts and Tajik literature. The dataset is derived from official materials prepared for the Centralized Entrance Exams (Markazi Millii Testi - NTC.tj) for the year 2025.

Dataset Summary

Language: Tajik (tg)
Source: National Testing Center (Markazi Millii Testi) - Entrance Exams 2025.
Size: 1575 multiple-choice questions.
Task: Multiple-Choice Question Answering.

How to Use

To load and use the dataset:

from datasets import load_dataset
dataset = load_dataset("f1rdavs/TajLiB")
dataset["grammar"]

Example Output

{
    "id": "TJ0001",
    "question": "Дар кадом калима зада дар ҳиҷои аввал меояд?",
    "option_a": "саҳро",
    "option_b": "имрӯз",
    "option_c": "берун",
    "option_d": "сӯзан",
    "answer": "B"
}

Citation

If you use TajLiB in your research, experiments, or publications, please cite the dataset using the citation information provided below.

@dataset{tajlib2025,
  title     = {TajLiB: Tajik Linguistic Benchmark},
  author    = {Firdavs Sultonov},
  year      = {2025},
  publisher = {HuggingFace Datasets},
  url       = {https://huggingface.co/datasets/f1rdavs/TajLiB}
}