Research map / Optimization / Implicit bias and generalization
Surrogate Gap Minimization Improves Sharpness-Aware Training
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 |
|---|---|---|---|
| Add GSAM's orthogonal ascent step to SAM | ViT-B/32 · ImageNet-v1 | top-1 accuracy (%): 73.6 -> 76.8 | Helped |
| Add GSAM's orthogonal ascent step to SAM | ViT-B/32 · ImageNet-Real | top-1 accuracy (%): 80.3 -> 82.7 | Helped |
| Add GSAM's orthogonal ascent step to SAM | ViT-B/32 · ImageNet-V2 | top-1 accuracy (%): 60.0 -> 63.0 | Helped |
| Add GSAM's orthogonal ascent step to SAM | ViT-B/32 · ImageNet-R | top-1 accuracy (%): 24.0 -> 25.1 | Helped |
| Add GSAM's orthogonal ascent step to SAM | ViT-B/32 · ImageNet-C | top-1 accuracy (%): 50.7 -> 51.7 | Helped |
| Add an ascent step to SAM (ablation) | Mixer-S/32 · ImageNet-v1 | top-1 accuracy | Helped |
| Replace the orthogonal ascent step with direct minimization of a weighted sum of perturbed loss and surrogate gap (ablation) | ViT-B/32 · ImageNet-v1 | top-1 accuracy | Hurt |
| Optimize the original loss and surrogate gap instead of perturbed loss and surrogate gap (ablation) | GSAM · ViT-B/32 | top-1 accuracy | No clear effect |
Closest papers by meaning
- A Universal Class of Sharpness-Aware Minimization Algorithms (ICML 2024 · similarity 0.91)
- Sharpness-Aware Minimization: General Analysis and Improved Rates (ICLR 2025 · similarity 0.91)
- Tilted Sharpness-Aware Minimization (ICML 2025 · similarity 0.90)
- Friendly Sharpness-Aware Minimization (CVPR 2024 · similarity 0.90)
- Sharpness-Aware Minimization Efficiently Selects Flatter Minima Late In Training (ICLR 2025 · similarity 0.90)
- Lookbehind-SAM: k steps back, 1 step forward (ICML 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.