README.md

metadata

annotations_creators:
  - machine-generated
language:
  - arz
license:
  - other
multilinguality:
  - monolingual
size_categories:
  - 10K<n<100K
source_datasets:
  - ehovy/race
task_categories:
  - multiple-choice
task_ids:
  - multiple-choice-qa
paperswithcode_id: race
pretty_name: RACE
dataset_info:
  - config_name: high
    features:
      - name: example_id
        dtype: string
      - name: article
        dtype: string
      - name: answer
        dtype: string
      - name: question
        dtype: string
      - name: options
        sequence: string
    splits:
      - name: train
        num_bytes: 125987908
        num_examples: 47136
      - name: train_latin
        num_bytes: 107057635
        num_examples: 56556
      - name: test
        num_bytes: 7656551
        num_examples: 2911
      - name: test_latin
        num_bytes: 5964966
        num_examples: 3192
      - name: validation
        num_bytes: 7427116
        num_examples: 2823
      - name: validation_latin
        num_bytes: 5846740
        num_examples: 3126
    download_size: 93875773
    dataset_size: 259940916
  - config_name: high_test
    features:
      - name: article
        dtype: string
      - name: problems
        list:
          - name: answer
            dtype: string
          - name: options
            sequence: string
          - name: question
            dtype: string
    splits:
      - name: test
        num_bytes: 3122475
        num_examples: 962
      - name: test_latin
        num_bytes: 2353536
        num_examples: 1009
    download_size: 3203997
    dataset_size: 5476011
  - config_name: middle
    features:
      - name: example_id
        dtype: string
      - name: article
        dtype: string
      - name: answer
        dtype: string
      - name: question
        dtype: string
      - name: options
        sequence: string
    splits:
      - name: train
        num_bytes: 35083982
        num_examples: 21262
      - name: train_latin
        num_bytes: 28630455
        num_examples: 24734
      - name: test
        num_bytes: 2209022
        num_examples: 1335
      - name: test_latin
        num_bytes: 1639110
        num_examples: 1384
      - name: validation
        num_bytes: 2194587
        num_examples: 1355
      - name: validation_ltn
        num_bytes: 1618485
        num_examples: 1394
    download_size: 23334452
    dataset_size: 71375641
  - config_name: middle_test
    features:
      - name: article
        dtype: string
      - name: problems
        list:
          - name: answer
            dtype: string
          - name: options
            sequence: string
          - name: question
            dtype: string
    splits:
      - name: test
        num_bytes: 812985
        num_examples: 352
      - name: test_latin
        num_bytes: 601753
        num_examples: 360
    download_size: 826233
    dataset_size: 1414738
configs:
  - config_name: high
    data_files:
      - split: train
        path: high/train-*
      - split: train_latin
        path: high/train_latin-*
      - split: test
        path: high/test-*
      - split: test_latin
        path: high/test_latin-*
      - split: validation
        path: high/validation-*
      - split: validation_latin
        path: high/validation_latin-*
  - config_name: high_test
    data_files:
      - split: test
        path: high_test/test-*
      - split: test_latin
        path: high_test/test_latin-*
  - config_name: middle
    data_files:
      - split: train
        path: middle/train-*
      - split: train_latin
        path: middle/train_latin-*
      - split: test
        path: middle/test-*
      - split: test_latin
        path: middle/test_latin-*
      - split: validation
        path: middle/validation-*
      - split: validation_ltn
        path: middle/validation_ltn-*
  - config_name: middle_test
    data_files:
      - split: test
        path: middle_test/test-*
      - split: test_latin
        path: middle_test/test_latin-*

Dataset Card for EgyptianRACE Medium and High (Arabic and Latin Script)

Table of Contents

• Table of Contents
• Dataset Description
  • Dataset Summary
  • Supported Tasks
  • Languages
• Dataset Structure
  • Data Instances
  • Data Splits
• Dataset Creation
  • Curation Rationale
  • Source Data
  • Annotations
  • Personal and Sensitive Information
• Considerations for Using the Data
  • Social Impact of Dataset
  • Discussion of Biases
  • Other Known Limitations
• Additional Information
  • Dataset Curators
  • Licensing Information
  • Citation Information

Dataset Description

Dataset Summary

RACE Medium and High in (Egyptian Arabic and Latin Script) is a reading comprehension benchmark sourced from high school English exams. It tests advanced reading, summarization, and inference skills using multiple-choice questions with 4 answer choices.

Supported Tasks

• Task Category: Multiple-choice question answering
• Task: Selecting the correct answer from a list of options

Languages

The dataset is available in Egyptian Arabic and Latin Script.

Data Splits

All datasets include a test split. Some also contain a development split for few-shot purposes.

Dataset Creation

Curation Rationale

To evaluate LLMs in Egyptian Arabic and its Latin-script form using established MCQ formats across varied domains.

Personal and Sensitive Information

No personal or sensitive information is included.

Considerations for Using the Data

Social Impact of Dataset

Supports the development of robust LLMs for underrepresented dialects and writing systems.

Discussion of Biases

May inherit translation model biases; dialect variation not exhaustively covered.

Other Known Limitations

• Limited to test splits
• Focused on selected subjects from original datasets

Additional Information

Dataset Curators

• MBZUAI-Paris team

Licensing Information

• MIT License

Citation Information