AI Paper Atlas

Research map / Time series / Foundation models for time series

GradientStabilizer: Fix the Norm, Not the Gradient

Tianjin Huang, Zhangyang Wang, Haotian Hu, Zhenyu Zhang, Gaojie Jin, Xiang Li, Li Shen, Jiaxing Shang, Tianlong Chen, Ke Li, Lu Liu, Qingsong Wen and 1 more

ICML 2026 · Time series › Foundation models for time series · approach: LLM-based time series forecasting

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add GradientStabilizer to AdamAdam · C4 FP4 training, LLaMA-350Mvalidation perplexity: 21.57 -> 18.89Improved
add GradientStabilizer to AdamWAdamW · C4 FP4 training, LLaMA-130Mvalidation perplexity: 28.72 -> 26.66Improved
add GradientStabilizer to AdamWAdamW · C4 FP16 training, LLaMA-130Mvalidation perplexity: 24.31 -> 23.14Improved
add Norm Clip to AdamAdam · C4 FP4 training, LLaMA-350Mvalidation perplexity: 21.57 -> 22.29Hurt
add AGC to AdamWAdamW · ImageNet-1K, ViT-Btop-1 accuracy: 79.3 -> 38.5Hurt
add GradientStabilizer to AdamWAdamW · ImageNet-1K, ViT-Btop-1 accuracy: 79.3 -> 79.6Improved
add GradientStabilizer to AdamAdam · ImageNet-1K, ViT-B, weight decay 1e-4top-1 accuracy: 59.3 -> 78.7Improved
add gradient clipping (Norm Clip) to AdamAdam · ImageNet-1K, ViT-B, weight decay 5e-4top-1 accuracy: 52.9 -> 20.6Hurt
add GradientStabilizer to Adam at weight decay 1e-4Adam · ImageNet-1K, ConvNeXt-T, weight decay 1e-4top-1 accuracy: 35.2 -> 78.8Improved
add GradientStabilizer to Adam-MiniAdam-Mini · C4 FP4 training, LLaMA-130Mvalidation perplexity: 31.66 -> 27.75Improved
add GradientStabilizer to LionLion · C4 FP4 training, LLaMA-350Mvalidation perplexity: 23.59 -> 22.75Improved
vary decay gamma1 in [0.5, 0.8] and gamma2 in [0.95, 0.99999] (ablation)GradientStabilizer · LLaMA-130M FP4 C4final validation loss: variation less than 0.02No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy