Research map / Label-efficient and robust learning / Domain adaptation and generalization
SWAD: Domain Generalization by Seeking Flat Minima
Junbum Cha, Sanghyuk Chun, Kyungjae Lee, Han-Cheol Cho, Seunghyun Park, Yunsung Lee, Sungrae Park
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 |
|---|---|---|---|
| apply SWAD on top of ERM | ERM · 5 DG benchmarks average | out-of-domain accuracy: 63.3 -> 66.9 | Improved |
| apply SWAD on top of CORAL | CORAL · 5 DG benchmarks average | out-of-domain accuracy: 64.5 -> 67.3 | Improved |
| apply SWAD on top of SAM | SAM · 5 DG benchmarks average | out-of-domain accuracy: 64.5 -> 65.5 | Improved |
| apply SWAD on top of CORAL | CORAL · VLCS | out-of-domain accuracy: 78.8 -> 78.9 | No clear effect |
| apply SWAD on top of SAM | SAM · VLCS | out-of-domain accuracy: 79.4 -> 78.5 | Hurt |
| remove dense sampling (interval 100 instead of 1) (ablation) | SWAD · PACS+VLCS average | out-of-domain accuracy: 82.2 -> 83.0 | Improved |
| remove overfit-aware end iteration (average to final iteration) (ablation) | SWAD · PACS+VLCS average | out-of-domain accuracy: 82.4 -> 83.0 | Improved |
| remove overfit-aware start and end search (fixed 4000-5000) (ablation) | SWAD · PACS+VLCS average | out-of-domain accuracy: 81.7 -> 83.0 | Improved |
| choose averaging range on validation set instead of overfit-aware (ablation) | SWAD · PACS+VLCS average | out-of-domain accuracy: 83.0 -> 82.4 | Hurt |
| constant learning rate instead of cyclic for SWA (ablation) | SWA · PACS+VLCS average | out-of-domain accuracy: 81.2 -> 81.6 | No clear effect |
| SWAD instead of SWA on ImageNet | ResNet-50 · ImageNet-C | mCE: 56.8 -> 55.7 | Improved |
| SWAD instead of ERM on background challenge | ResNet-50 · BGC | accuracy: 8.7 -> 11.8 | Improved |
Later papers that compared with this method
23 later accepted papers report SWAD as a baseline in their results tables; 21 of them report a better number on most of the results they share with it.
- DomED: Redesigning Ensemble Distillation for Domain Generalization (ICML 2026 · better on 2 of 2 shared results)
- Representation Enhancement-Stabilization: Reducing Bias-Variance of Domain Generalization (ECCV 2024 · better on 3 of 3 shared results)
- Domain Generalization with Small Data (IJCV 2024 · better on 3 of 3 shared results)
- Winning Prize Comes from Losing Tickets: Improve Invariant Learning by Exploring Variant Parameters for Out-of-Distribution Generalization (IJCV 2024 · better on 1 of 1 shared result)
- DART: Diversify-Aggregate-Repeat Training Improves Generalization of Neural Networks (CVPR 2023 · better on 6 of 6 shared results)
- POEM: Polarization of Embeddings for Domain-Invariant Representations (AAAI 2023 · better on 5 of 5 shared results)
- Domain Generalisation via Domain Adaptation: An Adversarial Fourier Amplitude Approach (ICLR 2023 · better on 5 of 5 shared results)
- Generalizable Decision Boundaries: Dualistic Meta-Learning for Open Set Domain Generalization (ICCV 2023 · better on 4 of 4 shared results)
15 more not listed.
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Seeking Consistent Flat Minima for Better Domain Generalization via Refining Loss Landscapes (CVPR 2025 · similarity 0.91)
- Gradient-Guided Annealing for Domain Generalization (CVPR 2025 · similarity 0.89)
- Unknown Domain Inconsistency Minimization for Domain Generalization (ICLR 2024 · similarity 0.89)
- One-Step Generalization Ratio Guided Optimization for Domain Generalization (ICML 2025 · similarity 0.88)
- PracticalDG: Perturbation Distillation on Vision-Language Models for Hybrid Domain Generalization (CVPR 2024 · similarity 0.87)
- Rethinking Multi-domain Generalization with A General Learning Objective (CVPR 2024 · similarity 0.87)
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.
