Research map / Label-efficient and robust learning / Domain adaptation and generalization
"On Multi-Domain Long-Tailed Recognition, Imbalanced Domain Generalization and Beyond"
Yuzhe Yang, Hao Wang, Dina Katabi
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 BoDA loss (decoupled classifier) | ERM · MDLT average over five datasets | accuracy: 77.6 -> 80.2 | Improved |
| add decoupled classifier learning stage with class-balanced sampling (ablation) | BoDA · MDLT average over five datasets | accuracy: 78.8 -> 80.2 | Improved |
| add momentum to BoDA (ablation) | BoDA_r,c · MDLT average over five datasets | accuracy: 80.2 -> 80.5 | Improved |
| use balanced distance (BoDA instead of vanilla DA) (ablation) | DA loss · MDLT datasets average | accuracy: 2.3% on average | Improved |
| vary distance calibration coefficient lambda (ablation) | BoDA · MDLT datasets average | accuracy: 1.9% to 2.9% over ERM | No clear effect |
| apply BoDA on severe-imbalance dataset | ERM · TerraInc-MLT | average accuracy: 75.3 -> 83.0 | Improved |
| apply BoDA on balanced-ish dataset | ERM · PACS-MLT | average accuracy: 97.1 -> 97.2 | No clear effect |
| apply BoDA on few-shot pairs | ERM · PACS-MLT few-shot pairs | accuracy: 98.0 -> 96.0 | Hurt |
| apply BoDA to zero-shot pairs | ERM · VLCS-MLT zero-shot pairs | accuracy: 32.9 -> 43.3 | Improved |
| add BoDA to domain generalization | ERM · DG average over five datasets | accuracy: 63.3 -> 65.5 | Improved |
| combine BoDA with current DG SOTA | Current SOTA · DG average over five datasets | accuracy: 64.5 -> 66.2 | Improved |
| use BoDA over TerraInc DG | ERM · TerraInc (DG) | accuracy: 46.1 -> 50.2 | Improved |
Closest papers by meaning
- Domain-Conditioned Class Imbalance: Why Global Class Balance Fails Across Domains (NeurIPS 2026 · similarity 0.90)
- Rethinking the Value of Labels for Improving Class-Imbalanced Learning (NeurIPS 2020 · similarity 0.90)
- Long-tailed Recognition with Model Rebalancing (NeurIPS 2025 · similarity 0.88)
- Difficulty-aware Balancing Margin Loss for Long-tailed Recognition (AAAI 2025 · similarity 0.88)
- Harnessing Hierarchical Label Distribution Variations in Test Agnostic Long-tail Recognition (ICML 2024 · similarity 0.88)
- Robust Visual Recognition with Class-Imbalanced Open-World Noisy Data (AAAI 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.
