Alternative EOI D - MENA Open Digital-Economy Benchmark
Title (paste-ready): MENA-Bench: An Open Benchmark Suite and Leaderboard for Digital-Economy AI in the Middle East and North Africa
EOI BODY (paste-ready)
The Middle East and North Africa contributes a growing share of global digital-economy activity. Tokenised real estate, Islamic-finance products, supply-chain corridors, and smart-city data all generate distinctive signals. Existing AI benchmarks evaluate models on US and European data. No curated benchmark suite covers MENA digital-economy tasks. The gap shapes which models and methods researchers choose to develop. A regional benchmark hosted at AUS would change that calculus.
We propose MENA-Bench, an open CC-BY benchmark suite plus public leaderboard for digital-economy AI in MENA. The suite covers six tasks across three families. Tasks one and two cover high-frequency cryptocurrency anomaly detection plus tokenised real-estate price impact analysis using Prypco and DLD pilot signals. Tasks three and four cover Islamic-finance smart-contract risk classification against AAOIFI rule encodings plus supply-chain shock forecasting on Gulf trade corridors. Tasks five and six cover sentiment classification on Arabic and English financial news drawing on Joerg’s 2025 reaction-times-to-news work plus macro-indicator nowcasting on UAE and regional time series. Each task ships with a labelled dataset, baseline models, and an evaluation harness. The methodological precedent comes from FRG23 anomaly detection at AUS plus FRG24 price-impact analysis. AR-Pro (NeurIPS 2024) provides the counterfactual-explanation reference for the anomaly task. Foundation-model tools serve inference baselines only with no model training in scope.
Specific aims: (1) Curate six labelled MENA digital-economy datasets. (2) Implement baselines and the leaderboard. (3) Release MENA-Bench under CC-BY year 1. (4) Attract at least ten external research-group submissions. (5) Train one UAE-based PhD plus host two workshops.
Application-readiness arc. Months 0-6 deliver the Applied baseline with the first three tasks curated and a leaderboard prototype. Months 6-12 add the remaining three tasks and a v0 release. Months 12-18 hit Tech Development with v1 and an external-submission window. Months 18-24 demonstrate the first Validation step. By month 24, at least ten external groups have submitted entries via the COST 19130 and MSCA-DN networks.
PI Stephen Chan is at AUS Mathematics & Statistics. Co-PI Joerg Osterrieder coordinates MSCA-DN Digital Finance and chairs COST Action 19130 across 51 countries. Co-PI Youcheng Sun at MBZUAI Computer Science leads verification of the leaderboard scoring and benchmark integrity. UAE delivery anchors at AUS plus MBZUAI through Co-PI presence. Two UAE-based PhDs are recruited, one at AUS under Stephen and one at MBZUAI under Youcheng. Capacity building includes the AUS course Interpretable AI for Tokenised Markets and workshops at AUS year 1 and MBZUAI year 2. ATRC is target outreach during the full-proposal phase. All code MIT, datasets CC-BY where licence allows.