Research map / Label-efficient and robust learning / OOD and anomaly detection
Balanced Hyperbolic Embeddings Are Natural Out-of-Distribution Detectors
Tejaswi Kasarla, Max van Spengler, Pascal Mettes
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 Euclidean head with balanced hyperbolic prototypes for MSP scoring | ResNet-34 MSP · CIFAR-100 (OpenOOD, avg over OOD sets) | FPR@95: 58.24 -> 49.46 | Improved |
| replace Euclidean head with balanced hyperbolic prototypes for SCALE scoring | ResNet-34 SCALE · CIFAR-100 (OpenOOD, avg over OOD sets) | AUROC: 79.68 -> 79.20 | Mixed |
| replace Euclidean head with balanced hyperbolic prototypes for KNN scoring | KNN · ImageNet100 (OpenOOD, avg over OOD sets) | FPR@95: 37.13 -> 45.74 | Hurt |
| use hyperbolic one-hot prototypes instead of Euclidean classifier (ablation) | Euclidean baseline · CIFAR-100 (5 scoring functions) | FPR@95 / AUROC: 4/5 (FPR@95) and 3/5 (AUROC) scoring functions | Improved |
| add hierarchical distortion objective and subhierarchy balancing on hyperbolic one-hot prototypes (ablation) | Hyperbolic embeddings · CIFAR-100 (5 scoring functions) | OOD metrics | Improved |
| add norm loss to distortion loss (ablation) | Balanced Hyperbolic Embeddings · toy imbalanced tree | norm variance | Improved |
| use curvature c=1 instead of 0.5 (ablation) | Balanced Hyperbolic · CIFAR-100 curvature ablation (MSP) | FPR@95: 73.40 -> 49.46 | Improved |
| use curvature c=1 instead of 2.0 (ablation) | Balanced Hyperbolic · CIFAR-100 curvature ablation (MSP) | FPR@95: 85.29 -> 49.46 | Improved |
| vary embedding dimension from 8 to 64 (ablation) | Balanced Hyperbolic · CIFAR-100 | MAP: 0.84 -> 0.88 | No clear effect |
| train with hierarchy via hyperbolic prototypes instead of Euclidean baseline | Euclidean baseline · CIFAR-100 OSR splits (hierarchical OOD, CIFAR-OOD split) | H-Dist / HSI-b1: 3.25 -> 2.32 H-Dist; 31.83 -> 67.21 HSI-b1 | Improved |
| balanced hyperbolic embeddings instead of Poincare Embeddings | MSP scoring · CIFAR-100 (MSP scoring, vs other hyperbolic) | FPR@95 / AUROC: 50.50 -> 49.46 FPR@95; 83.48 -> 82.43 AUROC | Mixed |
| balanced hierarchical hyperbolic prototypes instead of hyperspherical PALM prototypes | KNN · CIFAR-100 (KNN k=300, 128-dim prototypes) | FPR@95 / n-AUROC: 38.27 -> 35.83 FPR@95; 78.96 -> 78.50 n-AUROC | Improved |
Closest papers by meaning
- Hyperbolic Embeddings of Supervised Models (NeurIPS 2024 · similarity 0.89)
- Robust large-margin learning in hyperbolic space (NeurIPS 2020 · similarity 0.89)
- Learning Structured Representations with Hyperbolic Embeddings (NeurIPS 2024 · similarity 0.87)
- HYPO: Hyperspherical Out-Of-Distribution Generalization (ICLR 2024 · similarity 0.87)
- Hyperbolic Dataset Distillation (NeurIPS 2025 · similarity 0.87)
- Fully Hyperbolic Convolutional Neural Networks for Computer Vision (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.
