Research map / Label-efficient and robust learning / OOD and anomaly detection
FEVER-OOD: Free Energy Vulnerability Elimination for Robust Out-of-Distribution Detection
Brian K.S. Isaac-Medina, Mauricio Che, Yona Falinie A. Gaus, Samet Akcay, Toby P. Breckon
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 |
|---|---|---|---|
| Add null space reduction layer (96-NSR) plus LSV regulariser to VOS (ablation) | VOS · CIFAR-10 | FPR95: 33.66 -> 28.22 | Improved |
| Add 96-NSR plus LSV regulariser to FFS (ablation) | FFS · CIFAR-10 | AUROC: 90.52 -> 94.03 | Improved |
| Reduce features all the way to 10 dimensions (10-NSR) (ablation) | VOS · CIFAR-10 | FPR95: 33.66 -> 37.61 | Hurt |
| Add LSV regulariser only (no NSR) to VOS (ablation) | VOS · CIFAR-10 | FPR95: 33.66 -> 33.19 | No clear effect |
| Use condition number regulariser instead of LSV regulariser (ablation) | VOS · CIFAR-10 | AUROC: 92.92 (LSVR) vs 91.98 (CNR) | Hurt |
| Add 114-NSR plus LSV regulariser to VOS (ablation) | VOS · CIFAR-100 | FPR95: 71.01 -> 68.57 | Improved |
| Add 100-NSR plus LSV regulariser to Dream-OOD (ablation) | Dream-OOD · CIFAR-100 | FPR95: 50.96 -> 42.77 | Improved |
| Add 100-NSR plus LSV regulariser to Dream-OOD (ablation) | Dream-OOD · ImageNet-100 | FPR95: 39.33 -> 36.50 | Improved |
| Add LSV regulariser to pretrained Dream-OOD without NSR layer (ablation) | Dream-OOD · ImageNet-100 | FPR95: 39.33 -> 65.25 | Hurt |
| Use condition number regulariser with 100-NSR (ablation) | VOS, FFS, Dream-OOD · CIFAR-100 | convergence | Hurt |
| Add 768-NSR plus condition number regulariser to VOS detector | VOS (Faster R-CNN) · PASCAL VOC to MS-COCO | FPR95: 50.29 -> 47.47 | Improved |
| Add FEVER-OOD to object detectors | VOS/FFS · PASCAL VOC | mAP: 0.489 -> 0.491 | No clear effect |
Closest papers by meaning
- Dual Energy-Based Model with Open-World Uncertainty Estimation for Out-of-distribution Detection (CVPR 2025 · similarity 0.89)
- Is Complex Training Necessary for Long-Tailed OOD Detection? A Re-think from Feature Geometry (NeurIPS 2026 · similarity 0.88)
- MultiOOD: Scaling Out-of-Distribution Detection for Multiple Modalities (NeurIPS 2024 · similarity 0.88)
- The Best of Both Worlds: On the Dilemma of Out-of-distribution Detection (NeurIPS 2024 · similarity 0.88)
- Spurious-Aware Prototype Refinement for Reliable Out-of-Distribution Detection (NeurIPS 2025 · similarity 0.88)
- Position: Supervised Classifiers Answer the Wrong Questions for OOD Detection (ICML 2025 · similarity 0.88)
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.
