Research map / Architectures and efficiency / RNNs and normalization
Understanding and Improving Layer Normalization
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| replace LayerNorm bias and gain with input-adaptive scaling | LayerNorm · En-De | BLEU: 0.2 BLEU | Helped |
| replace LayerNorm bias and gain with input-adaptive scaling | LayerNorm · De-En | BLEU: 0.1 BLEU | Helped |
| replace LayerNorm bias and gain with input-adaptive scaling | LayerNorm · En-Vi | BLEU: 0.2 BLEU | Helped |
| replace LayerNorm bias and gain with input-adaptive scaling | LayerNorm · RT | accuracy: 0.29 ACC | Helped |
| replace LayerNorm bias and gain with input-adaptive scaling | LayerNorm · SST5 | accuracy: 1.31 ACC | Helped |
| replace LayerNorm bias and gain with input-adaptive scaling | LayerNorm · MNIST | accuracy: 0.22 ACC | Helped |
| replace LayerNorm bias and gain with input-adaptive scaling | LayerNorm · PTB | UAS: 0.11 UAC | Helped |
| remove LayerNorm bias and gain (ablation) | LayerNorm · En-Vi | BLEU: 0.4 BLEU | Helped |
| remove LayerNorm bias and gain (ablation) | LayerNorm · SST5 | accuracy: 1.31 ACC | Helped |
| remove LayerNorm bias and gain (ablation) | LayerNorm · En-De | BLEU | No clear effect |
| detach derivatives of mean and variance while retaining forward normalization (ablation) | LayerNorm-simple · De-En | BLEU: 1.6 | Hurt |
| detach derivatives of mean and variance while retaining forward normalization (ablation) | LayerNorm-simple · En-Vi | BLEU: 3.9 | Hurt |
Closest papers by meaning
- Impact of Layer Norm on Memorization and Generalization in Transformers (NeurIPS 2025 · similarity 0.88)
- The Curse of Depth in Large Language Models (NeurIPS 2025 · similarity 0.88)
- Mix-LN: Unleashing the Power of Deeper Layers by Combining Pre-LN and Post-LN (ICLR 2025 · similarity 0.87)
- HybridNorm: Towards Stable and Efficient Transformer Training via Hybrid Normalization (NeurIPS 2025 · similarity 0.87)
- Outlier Weighed Layerwise Sparsity (OWL): A Missing Secret Sauce for Pruning LLMs to High Sparsity (ICML 2024 · similarity 0.86)
- DLP: Dynamic Layerwise Pruning in Large Language Models (ICML 2025 · similarity 0.86)
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.