Tag
This paper investigates whether large language models treat questions about religious conversions symmetrically, finding persistent asymmetries with certain faiths favored over others. The study tests 20 models across 182 religious pairings, revealing reproducible patterns that could have real-world implications.