Research map / Label-efficient and robust learning / OOD and anomaly detection
CADet: Fully Self-Supervised Out-Of-Distribution Detection With Contrastive Learning
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 |
|---|---|---|---|
| use CADet with a self-supervised contrastive backbone instead of ViM with a supervised backbone | ImageNet-O detection · ImageNet-O | AUROC: 81.02 -> 82.29 | Helped |
| use CADet with a supervised backbone instead of the strongest listed supervised baseline | iNaturalist detection · iNaturalist | AUROC: 89.48 -> 95.28 | Helped |
| use self-supervised contrastive training rather than supervised training for CADet (ablation) | CADet · ImageNet-O | AUROC: 70.73 -> 82.29 | Helped |
| use self-supervised contrastive training rather than supervised training for CADet (ablation) | CADet · iNaturalist | AUROC: 95.28 -> 83.42 | Hurt |
| use self-supervised contrastive training rather than supervised training for CADet (ablation) | CADet · ImageNet adversarial detection | Average AUROC: 76.57 -> 96.12 | Helped |
| use CADet with a self-supervised contrastive backbone rather than Mahalanobis with a supervised backbone | ImageNet adversarial detection · PGD | AUROC: 92.71 -> 94.88 | Helped |
| add clean calibration to MMD using independent in-distribution samples (ablation) | MMD two-sample test with SimCLRv2 · PGD, two-sample test, 3 samples | AUROC: 35.2 -> 70.5 | Helped |
| add clean calibration to MMD using independent in-distribution samples (ablation) | MMD two-sample test with SimCLRv2 · CIFAR-10 vs CIFAR-10.1, n=200 | Rejection rate: 0.702 -> 0.686 | Hurt |
| combine within-sample and cross-sample similarities in CADet | CADet · ImageNet OOD and adversarial detection | AUROC | Helped |
| calibrate the weight of cross-sample similarity by equalizing validation variances rather than setting gamma to 1 | CADet · ImageNet OOD and adversarial detection | AUROC | Helped |
| use contrastive rather than supervised features with MMD (ablation) | MMD two-sample test · CIFAR-10 vs CIFAR-10.1, n=500 | Rejection rate: 0.884 -> 0.997 | Helped |
Closest papers by meaning
- Adaptive Multi-prompt Contrastive Network for Few-shot Out-of-distribution Detection (ICML 2025 · similarity 0.89)
- Learning with Mixture of Prototypes for Out-of-Distribution Detection (ICLR 2024 · similarity 0.87)
- Provable Discriminative Hyperspherical Embedding for Out-of-Distribution Detection (AAAI 2025 · similarity 0.86)
- Splitting & Integrating: Out-of-Distribution Detection via Adversarial Gradient Attribution (ICML 2025 · similarity 0.86)
- Image Background Serves as Good Proxy for Out-of-distribution Data (ICLR 2024 · similarity 0.86)
- Expecting The Unexpected: Towards Broad Out-Of-Distribution Detection (NeurIPS 2024 · similarity 0.86)
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.