Research map / Label-efficient and robust learning / Test-time adaptation
CODA: Generalizing to Open and Unseen Domains with Compaction and Disambiguation
Chaoqi Chen, Luyao Tang, Yue Huang, Xiaoguang Han, Yizhou Yu
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 source compaction (no target disambiguation) (ablation) | ERM · PACS | H-score: 36.1 -> 51.2 | Improved |
| add source compaction (ablation) | ERM · Office-Home | H-score: 34.3 -> 64.8 | Improved |
| add source compaction (ablation) | ERM · Digits | H-score: 18.0 -> 39.0 | Improved |
| add target disambiguation only (ablation) | ERM · PACS | H-score: 36.1 -> 49.8 | Improved |
| add target disambiguation on top of source compaction (ablation) | CODA · PACS | H-score: 51.2 -> 58.3 | Improved |
| add target disambiguation on top of source compaction (ablation) | CODA · Office-31 | H-score: 75.3 -> 81.0 | Improved |
| add target disambiguation on top of source compaction (ablation) | CODA · Digits | H-score: 39.0 -> 41.3 | Improved |
| combine Eq. 3 and Eq. 4 instead of Eq. 3 alone (ablation) | source compaction · PACS | H-score: 46.8 -> 51.2 | Improved |
| combine Eq. 3 and Eq. 4 instead of Eq. 3 alone (ablation) | source compaction · Digits | H-score: 38.6 -> 39.0 | Mixed |
| Eq. 2 objective instead of Eq. 1 (ablation) | unknown-aware training · PACS | H-score: 36.1 -> 41.1 | Improved |
| Eq. 3 virtual unknown objective instead of Eq. 1 (ablation) | unknown-aware training · Office-31 | H-score: 40.7 -> 72.6 | Improved |
| TAST test-time adaptation during testing | TAST · PACS | accuracy known vs unknown | Hurt |
Closest papers by meaning
- Bringing Equity to Classification: Domain Generalization for Domain-Linked Classes (TPAMI 2025 · similarity 0.90)
- Towards Generalizing to Unseen Domains with Few Labels (CVPR 2024 · similarity 0.89)
- Advancing Open-Set Domain Generalization Using Evidential Bi-Level Hardest Domain Scheduler (NeurIPS 2024 · similarity 0.89)
- Lost Domain Generalization Is a Natural Consequence of Lack of Training Domains (AAAI 2024 · similarity 0.89)
- Improving Domain Generalization with Domain Relations (ICLR 2024 · similarity 0.88)
- Set Valued Predictions For Robust Domain Generalization (ICML 2025 · similarity 0.88)
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.
