QIAS 2026 Shared Task: Second Call for Participation

11 views

Skip to first unread message

abdessalam.bouchekif

unread,

Jan 6, 2026, 6:30:45 AM (13 days ago) Jan 6

to SIGARAB: Special Interest Group on Arabic Natural Language Processing

We are pleased to invite the community to participate in QIAS 2026, a shared task on reasoning-oriented models for Islamic inheritance, organized as part of OSACT-7 and co-located with LREC 2026.

Overview
Do you want to build, test, and benchmark reasoning models?
QIAS 2026 is designed exactly for that purpose.

Following the success of QIAS 2025, we introduce QIAS 2026, a new edition focused on reasoning-centric models capable of understanding legal constraints, applying jurisprudential rules step by step, and generating verifiable intermediate reasoning.

Task Description
Given an Arabic natural-language description of an Islamic inheritance case—for example:

مات وترك: أم و أربع بنات ابن ابن. ما هو نصيب كل وريث؟

Systems must perform end-to-end Islamic inheritance reasoning by:

Identifying the eligible heirs, while distinguishing them from non-eligible relatives
Determining applicable blocking (ḥajb) relations
Assigning each eligible heir the correct juristic entitlement, including fixed shares and/or residuary shares
Applying key inheritance mechanisms such as ʿawl (proportional reduction) and radd (redistribution) when required
Producing a structured and verifiable intermediate output, together with the final numerical inheritance shares for each heir

The task explicitly evaluates a model’s ability to reason under Islamic legal constraints.

Dataset

We have developed a large-scale curated corpus of more than 10,000 Islamic inheritance cases to support QIAS 2026, each annotated with detailed step-by-step reasoning.

The dataset provides explicit intermediate stages—heir identification, blocking, legal entitlements, adjustment mechanisms, and final shares—enabling the development and evaluation of truly reasoning-centric models.

Why Participate

QIAS 2026 is ideal if you want to:

Develop reasoning-aware LLMs
Evaluate faithful, step-by-step legal reasoning
Benchmark models on structured intermediate outputs, not only final answers
Contribute to open and reproducible evaluation for Arabic legal reasoning
Push forward interpretable AI in sensitive legal and religious domains

Important Dates (Tentative)

December 21, 2025: Release of training, validation, and evaluation scripts
March 5, 2026: Test Set published
March 10, 2026: Final results released
March 25, 2026: System description paper submissions due
April 25, 2026: Camera-ready versions due
May 11-16, 2026: LREC 2026

How to Participate?

Register! Check all the details on the shared task Website (https://sites.google.com/view/qias2026)!
Join our Discussion Group (https://groups.google.com/g/qias2026) to follow the discussions and pose any questions about the task.
Participation as an academic, industrial, and independent researcher - everyone is welcome!

Venue

QIAS 2026 will be organized in conjunction with OSACT @ LREC 2026 (Spain).

May 11–16, 2026

https://sites.google.com/view/qias2026

Organizers

Abdessalam BOUCHEKIF (Hamad Bin Khalifa University, Qatar)
Samer RASHWANI (Hamad Bin Khalifa University, Qatar)
Mohammed GHALY (Hamad Bin Khalifa University, Qatar)
Mutaz A. AL-KHATIB (Hamad Bin Khalifa University, Qatar)
Emad Mohamed (Nazarbayev University, Kazakhstan )