吴正龙’s reinforcement-learning Bookmarks

LoRA may offer advantages in the cost and speed of post-training, and there are also a few operational reasons to prefer it to full fine-tuning.