Research map / Label-efficient and robust learning / OOD and anomaly detection
Likelihood Ratios for 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 |
|---|---|---|---|
| Add a perturbed-data background model and score by likelihood ratio rather than raw likelihood | PixelCNN++ · Fashion-MNIST vs. MNIST | AUROC: 0.089 (0.002) -> 0.994 (0.001) | Helped |
| Add L2 regularization to the perturbed-data background model (ablation) | PixelCNN++ likelihood-ratio method · Fashion-MNIST vs. MNIST | AUROC: 0.973 (0.031) -> 0.994 (0.001) | Helped |
| Add a perturbed-data background model and score by likelihood ratio rather than raw likelihood | PixelCNN++ · CIFAR-10 vs. SVHN | AUROC: 0.095 (0.003) -> 0.931 (0.032) | Helped |
| Add L2 regularization to the perturbed-data background model (ablation) | PixelCNN++ likelihood-ratio method · CIFAR-10 vs. SVHN | AUROC: 0.931 (0.032) -> 0.930 (0.042) | No clear effect |
| Add a perturbed-data background model and score by likelihood ratio rather than raw likelihood | LSTM · Genomic OOD test | AUROC: 0.626 (0.001) -> 0.755 (0.005) | Helped |
| Add L2 regularization to the perturbed-data background model (ablation) | LSTM likelihood-ratio method · Genomic OOD test | AUROC: 0.732 (0.015) -> 0.755 (0.005) | Helped |
| Rescale image inputs to the range -1 to 1 (ablation) | PixelCNN++ likelihood-ratio method · Fashion-MNIST vs. MNIST | AUROC: 0.994 (0.001) -> 0.936 (0.003) | Hurt |
| Increase the number of classifiers in the ensemble (ablation) | CNN ensemble · Genomic OOD test | AUROC: 0.682 (0.002) -> 0.690 (0.001) -> 0.695 (0.001) | Helped |
Closest papers by meaning
- A Geometric Explanation of the Likelihood OOD Detection Paradox (ICML 2024 · similarity 0.86)
- Out-of-Distribution Detection with Prototypical Outlier Proxy (AAAI 2025 · similarity 0.86)
- Enhancing the Power of OOD Detection via Sample-Aware Model Selection (CVPR 2024 · similarity 0.85)
- Learning with Mixture of Prototypes for Out-of-Distribution Detection (ICLR 2024 · similarity 0.85)
- Forte : Finding Outliers with Representation Typicality Estimation (ICLR 2025 · similarity 0.85)
- Expecting The Unexpected: Towards Broad Out-Of-Distribution Detection (NeurIPS 2024 · similarity 0.85)
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.