Research map / Label-efficient and robust learning / OOD and anomaly detection
GEN: Pushing the Limits of Softmax-Based Out-of-Distribution Detection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 Shannon entropy with truncated generalized entropy (GEN) | Softmax-based OOD scoring · ImageNet-1K ID; four OOD datasets averaged; Swin | AUROC: 89.60 -> 91.46 | Helped |
| Use GEN instead of Energy | Softmax-based OOD scoring · ImageNet-1K ID; four OOD datasets averaged; Swin | FPR95: 34.96 -> 32.28 | Helped |
| Add local feature clipping (ReAct*) to GEN (ablation) | GEN · ImageNet-1K ID; four OOD datasets averaged; RepVGG, ResNet-50-D, ViT, DeiT | AUROC: 85.75 -> 86.15 | Helped |
| Add local feature clipping (ReAct*) to GEN (ablation) | GEN · ImageNet-1K ID; four OOD datasets averaged; RepVGG, ResNet-50-D, ViT, DeiT | FPR95: 53.47 -> 53.48 | No clear effect |
| Add ReAct* feature clipping to GEN (ablation) | GEN · ImageNet-1K ID; four OOD datasets averaged; Swin | AUROC: 91.46 -> 91.45 | No clear effect |
| Add ReAct* feature clipping to GEN (ablation) | GEN · ImageNet-1K ID; four OOD datasets averaged; BiT-S-R101x1 | AUROC: 80.00 -> 80.22 | Helped |
| Add the feature-space residual score to GEN (ablation) | GEN · ImageNet-1K ID; four OOD datasets averaged; BiT-S-R101x1 | AUROC: 80.00 -> 91.61 | Helped |
| Add the feature-space residual score to GEN (ablation) | GEN · ImageNet-1K ID; four OOD datasets averaged; Swin | FPR95: 32.28 -> 31.83 | Helped |
| Use only the top M predicted class probabilities rather than summing over all classes (ablation) | GEN · ImageNet-1K OOD detection; ImageNet-O, iNaturalist, Textures, OpenImage-O | AUROC | Helped |
| Set the retained probability count to M = 100 (10% of classes) (ablation) | GEN · ImageNet-1K OOD detection; ImageNet-O, iNaturalist, Textures, OpenImage-O | AUROC | Helped |
| Lower the generalized-entropy exponent γ, using γ = 0.1 (ablation) | GEN · ImageNet-1K OOD detection; ImageNet-O, iNaturalist, Textures, OpenImage-O | AUROC | Helped |
Closest papers by meaning
- DOS: Diverse Outlier Sampling for Out-of-Distribution Detection (ICLR 2024 · similarity 0.87)
- Expecting The Unexpected: Towards Broad Out-Of-Distribution Detection (NeurIPS 2024 · similarity 0.87)
- Out-of-Distribution Detection with Prototypical Outlier Proxy (AAAI 2025 · similarity 0.86)
- Out-Of-Distribution Detection with Diversification (Provably) (NeurIPS 2024 · similarity 0.86)
- CASE: Exploiting Intra-class Compactness and Inter-class Separability of Feature Embeddings for Out-of-Distribution Detection (AAAI 2024 · similarity 0.86)
- Image Background Serves as Good Proxy for Out-of-distribution Data (ICLR 2024 · 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.