📣 Call for Participation — ⭐ HalluScoring 2026 ⭐ @ ArabicNLP 2026 | Islamic & General Culture QA
Bouchekif Abdessalam
Dear Colleagues,
We are pleased to invite you to participate in HalluScoring 2026, an ArabicNLP 2026 Shared Task on Hallucination Detection, Factuality Verification, and Trustworthy Arabic Question Answering across two domains: Islamic Knowledge and General Culture.
Official Website:
https://halluscoring.github.io/HalluScoring-2026/
HalluScoring 2026 aims to advance research on reliable Arabic Large Language Models that are not only fluent, but also factual, grounded, and trustworthy.
The shared task includes two main tracks:
Track 1 — Hallucination Detection
Detect hallucinations in Arabic question-answering systems using only question-answer pairs.
Task 1.1 — Generalize Across Questions: Detect hallucinations on unseen questions generated by the same set of LLMs.
Task 1.2 — Generalize Across Models: Detect hallucinations generated by unseen LLMs.
Track 2 — From Hallucination Detection to Truth
Participants must determine whether an answer is factual or hallucinated, and identify the correct answer from highly similar candidates.
Task 2.1 — Islamic Knowledge: Aqidah, Quran and Tafsir, Hadith and Sunnah, Fiqh, Islamic History, religious concepts, terminology, and authoritative Islamic sources.
Task 2.2 — General Culture: Geography, science, history, general knowledge, and Islamic culture.
Participants will have access to training and development datasets, baseline systems, starter kits, evaluation scripts, a submission platform, and full documentation.
We welcome students, researchers, NLP practitioners, machine learning engineers, academic teams, industry teams, and Arabic NLP enthusiasts.
Reliable AI starts with reliable evaluation. We look forward to your participation and innovative solutions.
Best regards,
HalluScoring 2026 Organizing Committee
Khaled Ziani
I want to know if it is possible to participate only in Task 2.1 — Islamic Knowledge, which seems similar to Shared Task 9: IslamicEval 2026!!! (for which we not yet have access to the data and the task description!!!!!!).
Thank you in advance.
Khaled Mohammed ZIANI