Research map / Label-efficient and robust learning / Domain adaptation and generalization
The Finer the Better: Towards Granular-aware Open-set Domain Generalization
Yunyun Wang, Zheng Duan, Xinyue Liao, Ke-Jia CHEN, Songcan Chen
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 semantic-aware prompt enhancement (SAPE) (ablation) | CLIP baseline · OSDG ablation | accuracy and H-score: +6.93% accuracy, +16.36% H-score | Improved |
| add semantic-guided diffusion generation (SGDG) (ablation) | CLIP baseline · OSDG ablation | accuracy and H-score: +4.13% accuracy, +9.67% H-score | Improved |
| add duplex contrastive learning (DCL) (ablation) | CLIP baseline · OSDG ablation | accuracy and H-score: +5.54% accuracy, +13.30% H-score | Improved |
| combine all three modules (SAPE, SGDG, DCL) (ablation) | CLIP baseline · Office-Home OSDG | H-score: 63.62 -> 98.97 | Improved |
| add alignment loss Lali (ablation) | SeeCLIP · OSDG ablation | accuracy and H-score: -2.89% accuracy, -3.15% H-score when removed | Improved |
| add repulsion loss Lrep (ablation) | SeeCLIP · OSDG ablation | accuracy and H-score: -1.54% accuracy, -1.60% H-score when removed | Improved |
| add cohesion loss Lcoh (ablation) | SeeCLIP · OSDG ablation | accuracy and H-score: -0.64% accuracy, -0.65% H-score when removed | Improved |
| add L1 regularization Lreg on semantic tokens (ablation) | SeeCLIP · PACS | accuracy: 99.5 -> 99.9 | Improved |
| use semantic-guided diffusion instead of OpenGAN, Cumix or ODG-CLIP for pseudo-unknowns (ablation) | SeeCLIP · Multi-Dataset OSDG | accuracy: 86.23 -> 88.70 vs ODG-CLIP generation | Improved |
| prompt learning methods without explicit unknown handling (CoOp, CoCoOp, MaPLe) | CLIP · OSDG average | H-score: 69.40 -> 41.65 (CoOp) | Hurt |
| use SeeCLIP in standard closed-set DG instead of ODG-CLIP | SeeCLIP · standard DG average | accuracy: 97.28 -> 97.78 | Improved |
Closest papers by meaning
- Unknown Prompt the only Lacuna: Unveiling CLIP's Potential for Open Domain Generalization (CVPR 2024 · similarity 0.93)
- OSLoPrompt: Bridging Low-Supervision Challenges and Open-Set Domain Generalization in CLIP (CVPR 2025 · similarity 0.92)
- Domain Generalization in CLIP via Learning with Diverse Text Prompts (CVPR 2025 · similarity 0.90)
- When and How Does CLIP Enable Domain and Compositional Generalization? (ICML 2025 · similarity 0.89)
- CLIPCEIL: Domain Generalization through CLIP via Channel rEfinement and Image-text aLignment (NeurIPS 2024 · similarity 0.89)
- Anchor-based Robust Finetuning of Vision-Language Models (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.
