OpenAI: Open-source RL-Teacher enables human-in-the-loop RL training for hard-to-specify rewards | SignalBreak | SignalBreak