January 25, 2026Fine-Tuning Language Models with Just Forward Passes零阶优化 & 收敛理论NoteOptimizationZeroth-OrderLLMTheoretical2023