Understanding Why Language Models Hallucinate: Testing Reasoning Against Priors · HackerLangs