Is there a primer on RL applied to LLMs?

Want to read more on how exactly new thinking models are trained and if some old RL techniques are now applied again to LLMs