AIBullisharXiv – CS AI · 5d ago7/102
🧠
GradientStabilizer:Fix the Norm, Not the Gradient
Researchers propose GradientStabilizer, a new technique to address training instability in deep learning by replacing gradient magnitude with statistically stabilized estimates while preserving direction. The method outperforms gradient clipping across multiple AI training scenarios including LLM pre-training, reinforcement learning, and computer vision tasks.