OPRIDE: Offline Preference-based Reinforcement Learning via In-Dataset Exploration

1. 1Identify the primary reasons for low query efficiency in offline PbRL, including inefficient exploration and lack of effective preference aggregation
2. 2Develop an in-dataset exploration approach to improve query efficiency
3. 3Implement OPRIDE, which leverages preference-based reinforcement learning to align with human intentions

ML researchers and engineers working on reinforcement learning and human-computer interaction can benefit from OPRIDE, as it addresses the challenge of low query efficiency in offline PbRL

💡 OPRIDE addresses the challenge of low query efficiency in offline PbRL by leveraging in-dataset exploration

🤖 OPRIDE improves offline PbRL query efficiency via in-dataset exploration!