I think you're seriously underestimating the importance of the RL steps on LLM performance. Also how do you think the most successful RL models have worked? Al

I think you're seriously underestimating the importance of the RL steps on LLM performance. Also how do you think the most successful RL models have worked? AlphaGo/AlphaZero both use Neural Networks for their policy and value networks which are the central mechanism of those models.