Inverse Preference Learning: Preference-based RL without a Reward Function

Feb-10-2026, 15:15:46 GMT–Neural Information Processing Systems 

Reward functions are difficult to design and often hard to align with human intent.