Research map / Label-efficient and robust learning / OOD and anomaly detection
Towards a Certificate of Trust: Task-Aware OOD Detection for Scientific AI
Bogdan Raonic, Siddhartha Mishra, Samuel Lanthaler
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 |
|---|---|---|---|
| use joint likelihood p(x,y_pred) as certificate instead of other joint-score certificates | JLBC · Average over 5 datasets | AUROC: 0.945 vs 0.927 (JDPath), 0.906, 0.925, 0.923 | Improved |
| use joint likelihood certificate on mixture-of-distributions Navier-Stokes | JLBC · NS-MIX | AUROC: 0.992 vs 0.918 (JDPath) | Improved |
| use joint likelihood certificate on MERRA2 humidity | JLBC · MERRA2 | ID/OOD accuracy: 0.956 vs 0.989 (JDPath) | Hurt |
| use joint likelihood certificate on brain segmentation | JLBC · Brain | AUROC: 0.808 vs 0.808 (JDPath) | No clear effect |
| use only input likelihood p(x) instead of joint p(x,y_pred) (ablation) | JLBC · NS-MIX | correlation with L1 error | Hurt |
| train the diffusion model for more epochs (up to 500) (ablation) | JLBC · Wave equation | likelihood-error alignment | Improved |
| use more decision samples from the training distribution (4 up to 32) (ablation) | JLBC · Wave equation | ID/OOD boundary stability | Improved |
| swap CNO regression model for ViT, UNet or C-FNO (ablation) | JLBC · Wave equation | likelihood-error correlation | No clear effect |
| use joint likelihood certificate instead of MC-Dropout and Rate-In | JLBC · Wave equation | accuracy and AUROC | Improved |
Closest papers by meaning
- MetaOOD: Automatic Selection of OOD Detection Models (ICLR 2025 · similarity 0.84)
- Beyond Clean Training Data: A Versatile and Model-Agnostic Framework for Out-of-Distribution Detection with Contaminated Training Data (CVPR 2025 · similarity 0.83)
- The Best of Both Worlds: On the Dilemma of Out-of-distribution Detection (NeurIPS 2024 · similarity 0.83)
- Fast Decision Boundary based Out-of-Distribution Detector (ICML 2024 · similarity 0.83)
- NECO: NEural Collapse Based Out-of-distribution detection (ICLR 2024 · similarity 0.83)
- Bridging OOD Detection and Generalization: A Graph-Theoretic View (NeurIPS 2024 · similarity 0.83)
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.
