Offline Reinforcement Learning of High-Quality Behaviors Under Robust Style Alignment | ScienceToStartup | ScienceToStartup