Tamim Sheikh

MIRAGE: Auditing Anti-Muslim Bias in Frontier LLMs Across Reasoning, Agentic, and Time-Coupled Conditions

MIRAGE (Muslim-Identity Reasoning and Agentic Generation Evaluation), a benchmark of 1,200 prompts spanning three deployment-realistic conditions, finds that chain-of-thought reasoning rather than suppresses Muslim-violence associations, suppressing direct-completion bias while leaving agentic asymmetry largely intact.