README / README.md
HassanB4's picture
Improve org README: clean layout, remove stale snapshot/leaderboard/methodology sections
905ffb8 verified
metadata
title: README
emoji: ๐Ÿ†
colorFrom: green
colorTo: blue
sdk: static
pinned: true

Haakkim โ€” ุญูŽูƒูู‘ู…

An open arena-style human preference evaluation platform for Arabic LLMs.

๐ŸŒ haakkim.tech  ยท  ๐Ÿ† Live Leaderboard  ยท  ๐Ÿ“ฆ Dataset


What is Haakkim?

Most Arabic LLM benchmarks rely on fixed tasks and MSA-only evaluation. Haakkim takes a different approach: real users compare two models side-by-side on real Arabic prompts, across 11 dialect varieties, and the results are aggregated into a statistically validated leaderboard.

Key features:

  • 11 Arabic varieties โ€” MSA, Saudi, Egyptian, Levantine, Tunisian, Iraqi, Moroccan, Algerian, Sudanese, Omani, Libyan
  • 3 evaluation modes โ€” Ranked Arena (official BT leaderboard), Side-by-Side, 10 Questions
  • Principled scoring โ€” Bradleyโ€“Terry with IPW sampling corrections and bootstrap CIs
  • Rankability gate โ€” BT scores only published when the comparison graph is fully connected and ESS is sufficient
  • Open & reproducible โ€” full dataset, audit logs, and scoring pipelines are public

Evaluation Modes

Mode Description Used for Ranking
Ranked Arena Random model pairing, single-turn MSA, matched system instruction โœ… Official BT leaderboard
Side-by-Side User-selected model pair, any dialect Win-rate only
10 Questions Fixed Arabic prompt pool, any dialect Win-rate only

Dataset

The first public release of Haakkim battle data is available on Hugging Face:

๐Ÿ“ฆ Haakkim/Haakkim-1.0v


Team

College of Computing, Umm Al-Qura University โ€” Mecca, Saudi Arabia

Name Role Email
Dr. Mourad Mars Principal Investigator msmars@uqu.edu.sa
Hassan Barmandah AI Researcher hassanhbarmandah@gmail.com
Abdulrhman Alassaf Software Engineer aaalassaf@outlook.com

Citation

@misc{mars2026haakkim,
  title        = {Haakkim: An Arena-Style Human Preference Evaluation Platform for Arabic {LLMs}},
  author       = {Mars, Mourad and Barmandah, Hassan and Alassaf, Abdulrhman},
  year         = {2026},
  howpublished = {\url{https://huggingface.co/datasets/Haakkim/Haakkim-1.0v}},
  note         = {College of Computing, Umm Al-Qura University, Mecca, Saudi Arabia}
}