Research map / Optimization / Implicit bias and generalization
Align-SAM: Seeking Flatter Minima for Better Cross-Subset Alignment
Van-Anh Nguyen, Mehrtash Harandi, Thanh-Toan Do, Linh Ngo Van, Dinh Phung, Trung Le
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add auxiliary-batch alignment (Align-SAM) | SAM ResNet34 · ImageNet from scratch | top-1 accuracy: 63.73 -> 65.89 | Improved |
| add auxiliary-batch alignment (Align-SAM) | SAM WRN28x10 · CIFAR-100 | accuracy: 83.00 -> 83.72 | Improved |
| add auxiliary-batch alignment (Align-ASAM) | ASAM WRN28x10 · CIFAR-100 | accuracy: 83.16 -> 83.88 | Improved |
| add auxiliary-batch alignment | SAM WRN28x10 · CIFAR-10 | accuracy: 96.87 -> 96.91 | No clear effect |
| add auxiliary-batch alignment (Align-ASAM) | ASAM WRN28x10 · CIFAR-10 | accuracy: 96.91 -> 97.15 | Improved |
| add auxiliary-batch alignment in transfer learning | SAM ResNet34 · ImageNet transfer | top-1 accuracy: 73.06 -> 73.94 | Improved |
| add auxiliary-batch alignment in transfer learning | SAM ViT-Adapter-S · ImageNet transfer | top-1 accuracy: 79.96 -> 80.04 | No clear effect |
| add auxiliary-batch alignment under 80% label noise | SAM ResNet32 · CIFAR-10 | accuracy: 61.69 -> 70.02 | Improved |
| add auxiliary-batch alignment under 80% label noise | SAM ResNet32 · CIFAR-100 | accuracy: 24.26 -> 27.66 | Improved |
| replace Sharp-MAML low with Align-SAM low | MAML · Omniglot 20-way 1-shot | accuracy: 92.89 -> 92.66 | Hurt |
| replace Sharp-MAML low with Align-SAM low | MAML · Mini-ImageNet 5-way 1-shot | accuracy: 49.72 -> 50.08 | Improved |
| increase auxiliary batch size from 16 to 128 (ablation) | Align-SAM ResNet32 · CIFAR-100 | accuracy: 70.58 -> 71.21 | Improved |
Closest papers by meaning
- Implicit Regularization of Sharpness-Aware Minimization for Scale-Invariant Problems (NeurIPS 2024 · similarity 0.92)
- Sharpness-Aware Minimization: General Analysis and Improved Rates (ICLR 2025 · similarity 0.92)
- Sharpness-Aware Minimization Enhances Feature Quality via Balanced Learning (ICLR 2024 · similarity 0.91)
- Sharpness-Aware Minimization Efficiently Selects Flatter Minima Late In Training (ICLR 2025 · similarity 0.91)
- Friendly Sharpness-Aware Minimization (CVPR 2024 · similarity 0.91)
- Changing the Training Data Distribution to Reduce Simplicity Bias Improves In-distribution Generalization (NeurIPS 2024 · similarity 0.90)
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.
