A new way to fine-tune LLMs just dropped
0:00 · Watch on YouTube
Evolution strategies are returning for language-model fine-tuning because 2025 work shows a population of about 30 parameter perturbations can improve billion-parameter models from a single whole-answer reward, and a...