Reinforcement Learning (RL) has long dominated fine tuning, powering many state of the art LLMs with methods like PPO and GRPO, which explore in action space. But what if we could explore directly in parameter space instead? Turns out, we can. A new framework proposes full parameter fine tuning using Evolution Strategies (ES), a gradient free approach optimizing directly in parameter space. By skipping gradients and exploring parameter space directly, Evolution Strategies (ES) achieve more accur