Research map / Label-efficient and robust learning / OOD and anomaly detection
Ranked Activation Shift for Post-hoc Out-of-Distribution Detection
Gianluca Guglielmo, Marc Masana
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 |
|---|---|---|---|
| replace ranked activations with ID reference profile (RAS) | EBO, ConvNeXt-T · ImageNet (ConvNeXt-T) | AUROC: 69.36 -> 83.31 | Improved |
| apply RAS instead of ASH-S/SCALE on non-rectified models | ViT-B/16 · ImageNet (ViT-B/16) | AUROC: 68.64 (SCALE) -> 81.48 | Improved |
| add RAS to EBO on ResNet50 | EBO · ImageNet (ResNet50) | AUROC: 84.03 -> 86.55 | Improved |
| add RAS to GEN | GEN · ImageNet (ResNet50) | AUROC: 84.60 -> 85.89 | Improved |
| add RAS to ViM* | ViM* · ImageNet (ResNet50) | AUROC: 84.29 -> 82.47 | Hurt |
| add RAS to ViM* | ViM* · CIFAR-100 | AUROC: 75.26 -> 75.28 | No clear effect |
| class-conditional reference profiles (ablation) | RAS · ImageNet | AUROC: 86.55 -> 84.39 | Hurt |
| shift only above-mean (inhibit) or below-mean (excite) activations (ablation) | RAS · OpenOOD | AUROC | Mixed |
| apply RAS to every residual block (RAS HL) (ablation) | RAS · OpenOOD | AUROC | Mixed |
| apply RAS to intermediate layers instead of penultimate (ablation) | RAS · CIFAR-10/100, ImageNet-200 | AUROC | No clear effect |
| add RAS | ID classifier · ImageNet (ResNet50) | top-1 accuracy: 76.17 -> 76.04 | No clear effect |
| add RAS vs AdaSCALE | EBO · ImageNet (ResNet50) | AUROC: 91.62 (AdaSCALE-A) vs 86.55 | 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.
- ReAct — ReAct: Out-of-distribution Detection With Rectified Activations (NeurIPS 2021) · better on 4 of 4 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
- Scaling for Training Time and Post-hoc Out-of-distribution Detection Enhancement (ICLR 2024 · similarity 0.89)
- Mysteries of the Deep: Role of Intermediate Representations in Out of Distribution Detection (NeurIPS 2025 · similarity 0.88)
- A Systematic Analysis of Out-of-Distribution Detection Under Representation and Training Paradigm Shifts (NeurIPS 2026 · similarity 0.87)
- Out-of-Distribution Detection by Leveraging Between-Layer Transformation Smoothness (ICLR 2024 · similarity 0.87)
- Discriminability-Driven Channel Selection for Out-of-Distribution Detection (CVPR 2024 · similarity 0.86)
- Splitting & Integrating: Out-of-Distribution Detection via Adversarial Gradient Attribution (ICML 2025 · 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.
