hypothetical behavior

off-policy data (1) safe exploration (1) human alignment (1) reward model (1) reward hacking (1) reward learning (1) value of information (1) reward function (1) trajectory optimization (1) inverse reinforcement learning (1)