GMTW-Ro: a deterministic benchmark for evaluating large language models on grounded Romanian tasks. Andrei-Stefan Bulzan, Bogdan Morariu, Andrei-Razvan Joldea, Diana Cernazanu-Glavan, Cosmin Cernazanu-Glavan. In: Frontiers in Artificial Intelligence, Vol. 9, 2 July 2026, DOI: 10.3389/frai.2026.1831918.