Research map / Label-efficient and robust learning / Semi-supervised learning
Adversarial Training Helps Transfer Learning via Better Representations
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 ℓ₂ adversarial training to source ImageNet training | ResNet-18 transfer to CIFAR-10 · CIFAR-10 (+0% pseudo-labels) | accuracy: 73.0% -> 82.8% | Helped |
| Add ℓ₂ adversarial training to source ImageNet training | ResNet-18 transfer to CIFAR-100 · CIFAR-100 (+0% pseudo-labels) | accuracy: 51.0% -> 62.6% | Helped |
| Add pseudo-labeled source examples | Standard ImageNet-trained ResNet-18 · CIFAR-10 | accuracy: 73.0% -> 78.8 % | Helped |
| Add pseudo-labeled source examples | Standard ImageNet-trained ResNet-18 · CIFAR-100 | accuracy: 51.0% -> 58.4% | Helped |
| Add pseudo-labeled source examples | ImageNet ResNet-18 with adversarial training · CIFAR-10 | accuracy: 82.8% -> 87.8 % | Helped |
| Add pseudo-labeled source examples | ImageNet ResNet-18 with adversarial training · CIFAR-100 | accuracy: 62.6% -> 69.5 % | Helped |
| Add ℓ₂ adversarial training to source ImageNet training | ResNet-18 transfer to Aircraft · Aircraft (+0% pseudo-labels) | accuracy: 17.3% -> 21.2% | Helped |
| Increase pseudo-label amount from +0% to +200% | ImageNet ResNet-18 with adversarial training · Aircraft | accuracy: 21.2% -> 20.9% | Hurt |
| Add ℓ₂ adversarial training to source ImageNet training | ResNet-18 transfer to Flowers · Flowers (+0% pseudo-labels) | accuracy: 60.7% -> 66.9% | Helped |
| Add ℓ₂ adversarial training to source ImageNet training | ResNet-18 transfer to Food · Food (+0% pseudo-labels) | accuracy: 33.7% -> 35.8% | Helped |
| Add ℓ₂ adversarial training to source ImageNet training | ResNet-18 transfer to Pets · Pets (+0% pseudo-labels) | accuracy: 43.2% -> 47.9% | Helped |
| Add pseudo-labeled source data | ImageNet-trained ResNet-18 · Flowers | accuracy: 60.7% -> 66.5% | Helped |
Closest papers by meaning
- Adversarially-Trained Deep Nets Transfer Better: Illustration on Image Classification (ICLR 2021 · similarity 0.92)
- Adversarial Training Reduces Information and Improves Transferability (AAAI 2021 · similarity 0.89)
- Boosting Adversarial Transferability through Augmentation in Hypothesis Space (CVPR 2025 · similarity 0.88)
- Initialization Matters for Adversarial Transfer Learning (CVPR 2024 · similarity 0.88)
- Learning to Transform Dynamically for Better Adversarial Transferability (CVPR 2024 · similarity 0.87)
- Enhancing Adversarial Transferability with Adversarial Weight Tuning (AAAI 2025 · 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.