“Risto Miikkulainen's research group has applied evolution strategies to fine-tune large language models such as GPT-4 and Llama after their initial pre-training.”