Research map / Label-efficient and robust learning / Noisy labels
Free Lunch for Domain Adversarial Training: Environment Label Smoothing
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 ELS to DANN | DANN · Office-31, ResNet18 | average accuracy: 81.3 -> 82.4 | Helped |
| Add ELS to SDAT | SDAT · Office-31, ResNet50 | average accuracy: 89.8 -> 90.4 | Helped |
| Add ELS to DANN | DANN · VLCS | average accuracy: 80.8 -> 81.5 | Helped |
| Add ELS to DANN | DANN · PACS | average accuracy: 86.1 -> 87.5 | Helped |
| Add ELS to SDAT | SDAT · Office-Home, ResNet-50 | average accuracy: 71.8 -> 72.6 | Helped |
| Add ELS to SDAT | SDAT · Office-Home, ViT | average accuracy: 84.3 -> 84.6 | Helped |
| Add ELS to DANN | DANN · Rotating MNIST | average accuracy: 52.1 -> 62.1 | Helped |
| Add ELS to DANN | DANN · CivilComments-Wilds | test worst-group accuracy: 61.7 ± 2.2 -> 66.0 ± 2.2 | Helped |
| Add ELS to DANN | DANN · RxRx1-Wilds | test ID accuracy: 22.9 ± 0.1 -> 26.7 ± 0.1 | Helped |
| Add ELS to DANN | DANN · Amazon-Wilds | test worst-group accuracy: 8.0 ± 0.0 -> 9.4 ± 0.0 | Helped |
| Use ELS with partially known or randomly assigned environment labels | DANN · PACS | accuracy | Helped |
| Gradually decrease the smoothing parameter during training | ELS-equipped DAT · PACS | generalization accuracy | Helped |
Closest papers by meaning
- Soften to Defend: Towards Adversarial Robustness via Self-Guided Label Refinement (CVPR 2024 · similarity 0.87)
- A Closer Look at Smoothness in Domain Adversarial Training (ICML 2022 · similarity 0.86)
- Training Deep Neural Networks with Virtual Smoothing Classes (AAAI 2025 · similarity 0.86)
- Kernel Learning with Adversarial Features: Numerical Efficiency and Adaptive Regularization (NeurIPS 2025 · similarity 0.85)
- Toward Improving Robustness and Accuracy in Unsupervised Domain Adaptation (AAAI 2025 · similarity 0.85)
- MaxSup: Overcoming Representation Collapse in Label Smoothing (NeurIPS 2025 · similarity 0.85)
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.