Dates: 2024
Principal Investigator: Dr. Irfan Ahmad

Description: This project investigates prompt engineering and machine translation to enhance instruction-following models in Arabic. It involves reviewing current methods, designing and testing various prompt templates (including prompt chaining and advanced techniques like CoT and self-ask), and collecting instruction samples from diverse domains. The study will compare manual expert prompts with automated ones, explore multi-turn dialogue prompts, and evaluate the role of machine-translated data in Arabic NLP. Performance will be assessed against state-of-the-art models using multiple metrics, and a supporting tool suite will be developed.