Learning Dynamics of LLM Finetuning | Xiaol.x | Podwise