Research map / Label-efficient and robust learning / OOD and anomaly detection
Unified Out-Of-Distribution Detection: A Model-Specific Perspective
Reza Averly, Wei-Lun Chao
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 |
|---|---|---|---|
| set detection threshold at TPR(ID+)=95 instead of TPR(ID)=95 (ablation) | OOD detection methods (MSP) · S-OOD detection, ResNet18 | FPR(S-OOD): SVHN 10.7 -> 2.5; Places 75.9 -> 51.0 | Improved |
| use MSP rather than advanced post-hoc scores for rejecting misclassified ID examples | Post-hoc OOD detection · ImageNet-1K ID-, ResNet50 | FPR(ID-)@TPR(ID+)=95: MSP 61.7 vs GradNorm 89.6 | Improved |
| stronger/more robust classifier improves detection of misclassified C-OOD | OOD detection · C-OOD (ImageNet-V2/S/R/A) | F1-score(C-OOD): MSP ImageNet-R: ResNet18 not stated vs ViT-B-16 78.9 | Improved |
| use more robust classifier (CLIP-ResNet50) when rejecting all C-OOD examples with MSP | C-OOD rejection · ImageNet-R | FPR(C-OOD): 71.6 -> 92.3 | Hurt |
| use robust ResNet50 (new training recipe) as deployed model | ID- and S-OOD detection · ImageNet-1K ID-, MSP | FPR(ID-): ResNet152 60.9 -> Robust ResNet50 66.5 despite higher ACC | Hurt |
| use GradNorm with stronger models | OOD detection · ID- and S-OOD | FPR: FPR(ID-) ResNet50 89.6 -> Robust ResNet50 99.2 | Hurt |
| increase ID accuracy of deployed classifier | OOD detection · S-OOD, mostly across methods | FPR(S-OOD) | Mixed |
| choose detection method per model for S-OOD | Post-hoc OOD scores · S-OOD datasets | FPR(S-OOD): ResNet50 SVHN ViM 0.2 vs MSP 4.2; ImageNet-O ViM 79.2 vs MSP 100.0 | Mixed |
Closest papers by meaning
- ImageNet-OOD: Deciphering Modern Out-of-Distribution Detection Algorithms (ICLR 2024 · similarity 0.93)
- Expecting The Unexpected: Towards Broad Out-Of-Distribution Detection (NeurIPS 2024 · similarity 0.92)
- The Best of Both Worlds: On the Dilemma of Out-of-distribution Detection (NeurIPS 2024 · similarity 0.92)
- Position: Supervised Classifiers Answer the Wrong Questions for OOD Detection (ICML 2025 · similarity 0.91)
- Image Background Serves as Good Proxy for Out-of-distribution Data (ICLR 2024 · similarity 0.91)
- Rethinking the Evaluation of Out-of-Distribution Detection: A Sorites Paradox (NeurIPS 2024 · similarity 0.91)
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.
