Research map / Label-efficient and robust learning / OOD and anomaly detection
Fisher-Rao Sensitivity for Out-of-Distribution Detection in Deep Neural Networks
Anthony Nguyen, Antonin Bertrand, Sylvie Le Hégarat-Mascle, Emanuel Aldea, Mohammed Nabil EL-KORSO, Renaud LUSTRAT, Franck FLORIN
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 additive product-manifold score instead of multiplicative Standard FIM Trace (ablation) | FIM Trace · ImageNet-O (ViT-B/16) | AUROC: 89.03 -> 92.61 | Improved |
| use additive score instead of multiplicative Standard FIM Trace (ablation) | FIM Trace · ImageNet-O (ResNet-50) | AUROC: 61.11 -> 71.77 | Improved |
| use additive score instead of multiplicative Standard FIM Trace (ablation) | FIM Trace · CIFAR-10 vs SVHN (ResNet-18) | AUROC: 86.16 -> 91.28 | Improved |
| use additive score instead of Tensor FIM Trace (ablation) | FIM Trace · ImageNet-O (ViT-B/16) | AUROC: 90.12 -> 92.61 | Improved |
| add tensor-subspace projection to Standard FIM Trace (ablation) | FIM Trace · ImageNet-O (ResNet-50) | AUROC: 61.11 -> 65.23 | Improved |
| use LDA instead of PCA for feature subspace (ablation) | Additive FIM Trace · ImageNet-1K OoD (ResNet-50) | AUROC combined: 76.73 -> 82.93 | Improved |
| use LDA subspace instead of no projection (ablation) | Additive FIM Trace · ImageNet-1K OoD (ResNet-50) | AUROC combined: 78.16 -> 82.93 | Improved |
| use LDA instead of PCA for residual signal (ablation) | Residual score · ImageNet-1K OoD (ResNet-50) | AUROC residual: 58.92 -> 80.23 | Improved |
| increase calibration set from 1,500 to 50,000 samples (ablation) | Additive FIM Trace · ImageNet-1K OoD (ResNet-50) | AUROC combined: 76.04 -> 82.97 | Improved |
| increase calibration set from 10,000 to 50,000 samples (ablation) | Additive FIM Trace · ImageNet-1K OoD (ResNet-50) | AUROC combined: 82.21 -> 82.97 | No clear effect |
| use Additive FIM Trace instead of ViM | Additive FIM Trace · ImageNet-O (ResNet-50) | AUROC: 68.21 -> 71.77 | Improved |
| use Additive FIM Trace instead of ViM | Additive FIM Trace · Places365 (ViT-B/16) | AUROC: 90.83 -> 89.51 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- GradOrth — GradOrth: A Simple yet Efficient Out-of-Distribution Detection with Orthogonal Projection of Gradients (NeurIPS 2023) · better on 5 of 8 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Igeood: An Information Geometry Approach to Out-of-Distribution Detection (ICLR 2022 · similarity 0.87)
- A Variational Information Theoretic Approach to Out-of-Distribution Detection (ICML 2025 · similarity 0.86)
- Improving Out-of-Distribution Detection via Dynamic Covariance Calibration (ICML 2025 · similarity 0.85)
- MahaVar: OOD Detection via Class-wise Mahalanobis Distance Variance under Neural Collapse (NeurIPS 2026 · similarity 0.85)
- Detecting Out-of-Distribution Through the Lens of Neural Collapse (CVPR 2025 · similarity 0.84)
- Trade-Offs of Diagonal Fisher Information Matrix Estimators (NeurIPS 2024 · similarity 0.84)
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.
