A comprehensive cognitive bias benchmark for evaluating Large Language Models, grounded in Kahneman-Tversky dual-process theory. Test 69 biases across 5 domains with 6 advanced metrics. Benchmarking 16 frontier models from 5 providers.