Research map / Probabilistic and scientific modeling / Variational and generative sampling
Hierarchical VAEs Know What They Don’t Know
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 the raw ELBO score with the likelihood ratio LLR^>1 (ablation) | HVAE · FashionMNIST (in) / MNIST (out) | AUROC: 0.268 -> 0.964 | Helped |
| Replace the L^>2 likelihood bound with LLR^>1 (ablation) | HVAE · FashionMNIST (in) / MNIST (out) | AUROC: 0.712 -> 0.964 | Helped |
| Use 250 importance samples for the likelihood ratio (ablation) | HVAE · FashionMNIST (in) / MNIST (out) | AUROC: 0.964 -> 0.984 | Helped |
| Use 250 importance samples for the likelihood ratio (ablation) | HVAE · FashionMNIST (in) / MNIST (out) | FPR80: 0.036 -> 0.013 | Helped |
| Use the L^>1 bound instead of the regular ELBO (ablation) | HVAE · FashionMNIST (in) / MNIST (out) | AUROC: 0.268 -> 0.593 | Helped |
| Use the L^>2 bound instead of the regular ELBO (ablation) | HVAE · FashionMNIST (in) / MNIST (out) | AUROC: 0.268 -> 0.712 | Helped |
| Use BIVA instead of the bottom-up HVAE (ablation) | Likelihood-ratio OOD detection · CIFAR10 (in) / SVHN (out) | AUROC: 0.811 -> 0.891 | Helped |
| Use the regular likelihood for OOD scoring (ablation) | HVAE · FashionMNIST (in) / MNIST (out) | AUROC: 0.964 -> 0.268 | Hurt |
Closest papers by meaning
- A Geometric Explanation of the Likelihood OOD Detection Paradox (ICML 2024 · similarity 0.90)
- Understanding Likelihood of Normalizing Flow and Image Complexity through the Lens of Out-of-Distribution Detection (AAAI 2024 · similarity 0.89)
- Detecting Out-of-Distribution Through the Lens of Neural Collapse (CVPR 2025 · similarity 0.88)
- Forte : Finding Outliers with Representation Typicality Estimation (ICLR 2025 · similarity 0.88)
- Mysteries of the Deep: Role of Intermediate Representations in Out of Distribution Detection (NeurIPS 2025 · similarity 0.88)
- Image Background Serves as Good Proxy for Out-of-distribution Data (ICLR 2024 · similarity 0.87)
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.