[mountainriver]

Love it! Its seems like a lot of the ideas from reinforcement learning are making their way into transformer land and NLP