[Jax_Hax]

There are models trained via RNN, but LLMs usually use Transformer architecture with a bit of human feedback on top which sort of uses reinforcement learning like AlphaGo