ADMD / README.md
riotu-lab's picture
Update README.md
c859706 verified
metadata
license: mit
language:
  - ar

Arabic Depth Mini Dataset (ADMD)

The Arabic Depth Mini Dataset (ADMD) is a compact yet highly challenging dataset designed to evaluate Arabic language models across diverse domains. It consists of 490 carefully curated questions spanning 10 major domains and 42 topics. The dataset is manually reviewed and emphasizes linguistic accuracy, cultural alignment, and methodological rigor.

Dataset Overview

Categories and Topics

The dataset covers 42 topics across 10 major domains:

1. Applied Sciences & Engineering

  • Mechanical Engineering
  • Computer Science
  • Medicine
  • Nutrition (includes Health & Fitness)
  • Earth Science

2. Natural Sciences

  • Biology
  • Cosmology

3. Social Sciences & Humanities

  • Psychology
  • Sociology
  • Anthropology
  • Media & Communication
  • Economics

4. Islamic & Religious Studies

  • Quranic Exegesis (Tafsir)
  • Hadith
  • Hadith Terminology (Mustalah)
  • Islamic Jurisprudence (Fiqh)
  • Principles of Islamic Jurisprudence (Usul Al-Fiqh)
  • Inheritance Laws (Fara'id)
  • Islamic Creed (Aqeedah)
  • Quranic Recitation Rules (Tajweed)
  • Quranic Readings (Qira’at)
  • Biography of the Prophet (Seerah)
  • Biographies of Islamic Scholars (Tarajim Al-Rijal)

5. Linguistics & Literature

  • Arabic Grammar (Nahw)
  • Arabic Morphology (Sarf)
  • Rhetoric (Balagha)
  • Arabic Prosody & Poetic Meters (Arood & Qawafi)
  • Poetry
  • Arabic Language
  • General Linguistics
  • Arabic Linguistics
  • Arabic Logic

6. Philosophy & Logic

  • Philosophy
  • Arabic Logic

7. Culture & Arts

  • Music
  • Folklore & Cultural Studies

8. Mathematics & Computational Sciences

  • Mathematics
  • Machine Learning

9. General & Miscellaneous Sciences

  • General Sciences
  • Cooking

10. Historical & Genealogical Studies

  • Genealogy (Ansab)

Model Evaluation Results

The dataset was used to evaluate five leading language models:

  1. GPT-4
  2. Claude 3.5 Sonnet
  3. Gemini Flash 1.5
  4. CommandR 100B
  5. Qwen-Max 2.5

Performance Metrics

The evaluation was conducted manually, and the results were categorized as follows:

  • True (T): The model answered correctly.
  • False (F): The model answered incorrectly.
  • Partially-True (PT): The model answered 60-80% correctly.
  • Partially-False (PF): The model answered 20-30% correctly.

Overall Model Performance

Model True (%) False (%) Partially-True (%) Partially-False (%)
Claude 3.5 Sonnet 33.5 43.5 18.2 4.8
CommandR 100B 15.0 54.0 15.6 15.4
Gemini Flash 1.5 22.1 56.2 12.0 9.7
GPT-4 11.8 67.3 17.3 3.5
Qwen-Max 2.5 13.1 57.4 17.8 11.7

Detailed Results by Domain

For detailed results by domain, refer to the tables below:

  1. Claude 3.5 Sonnet Results
  2. GPT-4 Results
  3. Gemini Flash 1.5 Results
  4. CommandR 100B Results
  5. Qwen-Max 2.5 Results

Claude 3.5 Sonnet Results

Field of Study True (%) False (%) Partially-True (%) Partially-False (%)
Applied Sciences & Engineering 42.00 28.00 24.00 6.00
Natural Sciences 45.00 5.00 45.00 5.00
Social Sciences & Humanities 38.00 38.00 20.00 4.00
Islamic & Religious Studies 30.00 41.82 16.36 11.82
Linguistics & Literature 12.84 66.97 13.76 6.42
Philosophy & Logic 50.00 50.00 0.00 0.00
Culture & Arts 15.00 65.00 15.00 5.00
Mathematics & Computational Sci. 50.00 20.00 20.00 10.00
General & Miscellaneous Sciences 51.67 40.00 8.33 0.00
Historical & Genealogical Studies 0.00 80.00 20.00 0.00

GPT-4 Results

Field of Study True (%) False (%) Partially-True (%) Partially-False (%)
Applied Sciences & Engineering 22.00 42.00 28.00 8.00
Natural Sciences 20.00 35.00 45.00 0.00
Social Sciences & Humanities 12.00 56.00 26.00 6.00
Islamic & Religious Studies 0.91 80.91 10.00 8.18
Linguistics & Literature 1.82 94.55 2.73 0.91
Philosophy & Logic 10.00 80.00 10.00 0.00
Culture & Arts 10.00 75.00 10.00 5.00
Mathematics & Computational Sci. 25.00 45.00 25.00 5.00
General & Miscellaneous Sciences 16.67 65.00 16.67 1.67
Historical & Genealogical Studies 0.00 100.00 0.00 0.00

Gemini Flash 1.5 Results

Field of Study True (%) False (%) Partially-True (%) Partially-False (%)
Applied Sciences & Engineering 24.00 46.00 24.00 6.00
Natural Sciences 40.00 15.00 20.00 25.00
Social Sciences & Humanities 38.00 32.00 14.00 16.00
Islamic & Religious Studies 0.00 88.18 5.45 6.36
Linguistics & Literature 2.75 84.40 4.59 8.26
Philosophy & Logic 15.00 60.00 5.00 20.00
Culture & Arts 10.00 70.00 10.00 10.00
Mathematics & Computational Sci. 45.00 30.00 25.00 0.00
General & Miscellaneous Sciences 36.67 56.67 1.67 5.00
Historical & Genealogical Studies 10.00 80.00 10.00 0.00

CommandR 100B Results

Field of Study True (%) False (%) Partially-True (%) Partially-False (%)
Applied Sciences & Engineering 30.00 52.00 6.00 12.00
Natural Sciences 30.00 15.00 50.00 5.00
Social Sciences & Humanities 18.00 46.00 20.00 16.00
Islamic & Religious Studies 3.64 69.09 10.00 17.27
Linguistics & Literature 4.55 82.73 3.64 9.09
Philosophy & Logic 10.00 45.00 15.00 30.00
Culture & Arts 15.00 70.00 5.00 10.00
Mathematics & Computational Sci. 25.00 30.00 25.00 20.00
General & Miscellaneous Sciences 13.33 60.00 11.67 15.00
Historical & Genealogical Studies 0.00 70.00 10.00 20.00

Qwen-Max 2.5 Results

Field of Study True (%) False (%) Partially-True (%) Partially-False (%)
Applied Sciences & Engineering 20.00 42.00 18.00 20.00
Natural Sciences 20.00 15.00 40.00 25.00
Social Sciences & Humanities 18.00 42.00 24.00 16.00
Islamic & Religious Studies 4.55 80.00 5.45 10.00
Linguistics & Literature 1.82 90.00 3.64 4.55
Philosophy & Logic 15.00 70.00 5.00 10.00
Culture & Arts 5.00 85.00 10.00 0.00
Mathematics & Computational Sci. 20.00 30.00 35.00 15.00
General & Miscellaneous Sciences 26.67 50.00 16.67 6.67
Historical & Genealogical Studies 0.00 70.00 20.00 10.00

and for the dataset here are the splits names

English Name Arabic Translation
machine_learning تعلم الآلة
Tafsir التفسير
Grammar_Nahw النحو
psycology علم النفس
nutrition التغذية
dialects اللهجات
mechanic الميكانيكا
biology علم الأحياء
cosmology علم الكون
Tarajem التراجم
Morphology_Sarf الصرف
Tajweed التجويد
Balaghah البلاغة
Orod_qawafi العروض والقوافي
Fiqh الفقه
Inheritance_mawarith المواريث
Osol_fiqh أصول الفقه
believe_Aqidah العقيدة
Mostalah المصطلح
Hadith الحديث
Islamic_History التاريخ الإسلامي
Maths الرياضيات
linguistics اللسانيات
Arabic_logic المنطق العربي
Ansab الأنساب
poety الشعر
Qiraat القراءات
old_music الموسيقى القديمة
culture الثقافة
philosophy الفلسفة
geology علم الأرض
language اللغة
economics الاقتصاد
sociology علم الاجتماع
medicine الطب
computer_science علوم الحاسوب
Anthropology علم الإنسان
media_com الإعلام والاتصال
general_language اللغة العامة
general_science العلوم العامة
cooking الطهي

Citation

If you use this dataset in your research, please cite the following paper:

@article{yourcitationkey,
  title={Towards a Quality Arabic Evaluation Datasets: Theory, Current Practice, and Future Directions},
  author={Your Name and Co-authors},
  journal={Your Journal},
  year={2024},
  publisher={Your Publisher}
}