Research map / Architectures and efficiency / RNNs and normalization
Sharpness-Aware Training for Free
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 SAM with SAF (trajectory loss, no extra gradient pass) | ResNet-50 · ImageNet | accuracy: 76.9 -> 77.8 | Helped |
| replace SAM with SAF | ResNet-101 · ImageNet | accuracy: 78.6 -> 79.3 | Helped |
| replace SAM with SAF | ViT-S/32 · ImageNet | accuracy: 68.9 -> 69.5 | Helped |
| replace SAM with SAF | ResNet-18 · CIFAR-10 | accuracy: 96.50 -> 96.37 | Hurt |
| replace SAM with SAF | PyramidNet-110 · CIFAR-100 | accuracy: 84.61 -> 84.71 | No clear effect |
| add SAF to SGD training | ResNet-50 · ImageNet | accuracy: 76.0 -> 77.8 | Helped |
| use EMA-model targets (MESA) instead of stored past outputs (SAF) | SAF · ImageNet | extra memory: 14,643 MB -> 98 MB | Helped |
| use EMA-model targets (MESA) instead of stored past outputs (SAF) | SAF · ImageNet ResNet-50 | training speed vs SGD: 99.1% -> 85.2% | Hurt |
Closest papers by meaning
- Sharpness-Aware Minimization: General Analysis and Improved Rates (ICLR 2025 · similarity 0.92)
- Friendly Sharpness-Aware Minimization (CVPR 2024 · similarity 0.92)
- Sharpness-Aware Minimization Efficiently Selects Flatter Minima Late In Training (ICLR 2025 · similarity 0.91)
- A Universal Class of Sharpness-Aware Minimization Algorithms (ICML 2024 · similarity 0.91)
- Sharpness-Aware Minimization Enhances Feature Quality via Balanced Learning (ICLR 2024 · similarity 0.90)
- A Single-Step, Sharpness-Aware Minimization is All You Need to Achieve Efficient and Accurate Sparse Training (NeurIPS 2024 · similarity 0.89)
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.