Researchers from Nanjing University and Carnegie Mellon University have introduced an AI approach that improves how machines learn from past data—a process known as offline reinforcement learning.