Evaluation awareness and faking, AI value trade-offs, misalignment propensity, reward hacking, CoT monitoring, training against lie detectors, and exploring the landscape of refusals