Research map / Label-efficient and robust learning / OOD and anomaly detection
Out-of-Distribution Detection using Multiple Semantic Label Representations
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 softmax supervision with multiple semantic embedding targets and use output norms for detection | CIFAR-10-trained ResNet-18 · SVHN, LSUN, CIFAR-100 OOD detection | AUROC: 99.48 vs 98.65; 96.05 vs 91.28; 91.23 vs 85.59 | Helped |
| replace softmax supervision with semantic embedding targets | CIFAR-100-trained ResNet-34 · SVHN OOD detection | AUROC: 87.42 vs 79.86 | Helped |
| replace softmax supervision with semantic embedding targets | CIFAR-100-trained ResNet-34 · CIFAR-10 OOD detection | AUROC: 75.99 vs 81.49 | Hurt |
| replace softmax supervision with semantic embedding targets | CIFAR-100-trained ResNet-34 · LSUN OOD detection | AUROC: 76.41 vs 79.05 | Hurt |
| replace softmax supervision with semantic embedding targets | LeNet · SC-23 → SC-7 | AUROC: 87.34 vs 85.99 | Helped |
| use 5 semantic predictors and disagreement to flag adversarial examples | DenseNet-121 · ImageNet adversarial examples | detection rate: 43.88% -> 62.04% | Helped |
| use 5 semantic predictors instead of the softmax ensemble at fixed false rejection rate | DenseNet-121 · ImageNet adversarial examples | detection rate: 15.41% -> 28.64% | Helped |
| increase embedding targets from 1 to 5 (ablation) | LeNet · GCommands | accuracy: 90.42 -> 91.13 | Helped |
| increase embedding targets from 1 to 5 (ablation) | ResNet-18 · CIFAR-10 | accuracy: 95.11 -> 95.04 | Hurt |
| increase embedding targets from 1 to 5 (ablation) | ResNet-34 · CIFAR-100 | accuracy: 77.62 -> 78.23 | Helped |
Closest papers by meaning
- Out-of-Distribution Detection via Deep Multi-Comprehension Ensemble (ICML 2024 · similarity 0.85)
- ProHOC: Probabilistic Hierarchical Out-of-Distribution Classification via Multi-Depth Networks (CVPR 2025 · similarity 0.85)
- Diffusion-based Layer-wise Semantic Reconstruction for Unsupervised Out-of-Distribution Detection (NeurIPS 2024 · similarity 0.84)
- Conjugated Semantic Pool Improves OOD Detection with Pre-trained Vision-Language Models (NeurIPS 2024 · similarity 0.84)
- Mysteries of the Deep: Role of Intermediate Representations in Out of Distribution Detection (NeurIPS 2025 · similarity 0.84)
- Splitting & Integrating: Out-of-Distribution Detection via Adversarial Gradient Attribution (ICML 2025 · 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.