THIQAH Company, in collaboration with King Abdullah University of Science and Technology (KAUST), presented a pioneering research paper at the Arabic Natural Language Processing Conference (ArabicNLP 2025), held in Suzhou, China, in conjunction with EMNLP 2025, one of the world’s leading conferences in artificial intelligence and language processing.

The paper, titled “ALARB – An Arabic Legal Argument Reasoning Benchmark,” introduces the first Arabic dataset designed to evaluate the legal reasoning capabilities of large language models. The study addresses the current lack of Arabic-specific benchmarks in the field of legal text processing—an area that has long been dominated by English-language resources—marking a major step toward enhancing Arabic’s presence in global AI research and innovation.

During the conference, THIQAH and KAUST researchers presented the dataset’s methodology, which involved collecting, organizing, and analyzing publicly available legal cases. These were then restructured into four main components — facts, reasoning, verdicts, and related regulations — allowing AI models to better understand and process Arabic legal texts across multiple reasoning tasks.

Initial results demonstrated notable improvements in model accuracy when tested on global systems such as GPT-4, Gemma, and Aya, highlighting the effectiveness of the ALARB benchmark in setting a new standard for evaluating Arabic AI models in legal reasoning.

This participation reinforces THIQAH’s commitment to supporting Arabic language technologies, advancing AI applications in legal and technical fields, and strengthening its international presence through collaboration with leading research institutions such as KAUST.

The ArabicNLP 2025 Conference is the world’s premier academic event dedicated to Arabic natural language processing, bringing together distinguished experts and researchers to explore the latest advancements in AI and its applications for Arabic digital content and linguistic innovation.