Research map / Label-efficient and robust learning / OOD and anomaly detection
Shaping Parameter Contribution Patterns for Out-of-Distribution Detection
Haonan Xu, Yang Yang
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 |
|---|---|---|---|
| add SPCP contribution truncation in training | Energy score, ResNet-18 · CIFAR-10 near-OOD | FPR95: 61.34 -> 31.67 | Improved |
| add SPCP contribution truncation in training | Energy score, ResNet-18 · CIFAR-10 far-OOD | FPR95: 41.69 -> 20.54 | Improved |
| add SPCP on CIFAR-100 | Energy score, ResNet-18 · CIFAR-100 near-OOD | FPR95: 55.62 -> 55.21 | No clear effect |
| add SPCP on CIFAR-100 | Energy score, ResNet-18 · CIFAR-100 far-OOD | FPR95: 56.59 -> 50.59 | Improved |
| add SPCP on ImageNet-200 | Energy score, ResNet-18 · ImageNet-200 near-OOD | FPR95: 60.24 -> 59.77 | No clear effect |
| add SPCP on ImageNet-200 | Energy score, ResNet-18 · ImageNet-200 far-OOD | FPR95: 34.86 -> 30.43 | Improved |
| truncate contributions only at inference (no training) (ablation) | SPCP · CIFAR-100 near-OOD | FPR95: 55.62 -> 60.28 | Hurt |
| truncate contributions only at inference (no training) (ablation) | SPCP · CIFAR-100 far-OOD | FPR95: 56.59 -> 53.66 | Improved |
| truncate during both training and inference instead of training only (ablation) | SPCP · CIFAR-100 far-OOD | FPR95: 51.96 -> 50.59 | Improved |
| disable dynamic threshold updating (EMA beta = 1.0) (ablation) | SPCP · CIFAR-100 | AUROC | Hurt |
| use very small initial threshold lambda0 (ablation) | SPCP · CIFAR-100 | AUROC | Hurt |
| use too small percentile rho (ablation) | SPCP · CIFAR-100 | AUROC | Hurt |
Closest papers by meaning
- Towards Optimal Feature-Shaping Methods for Out-of-Distribution Detection (ICLR 2024 · similarity 0.89)
- Splitting & Integrating: Out-of-Distribution Detection via Adversarial Gradient Attribution (ICML 2025 · similarity 0.88)
- Expecting The Unexpected: Towards Broad Out-Of-Distribution Detection (NeurIPS 2024 · similarity 0.88)
- Scaling for Training Time and Post-hoc Out-of-distribution Detection Enhancement (ICLR 2024 · similarity 0.88)
- A Variational Information Theoretic Approach to Out-of-Distribution Detection (ICML 2025 · similarity 0.88)
- Mysteries of the Deep: Role of Intermediate Representations in Out of Distribution Detection (NeurIPS 2025 · similarity 0.88)
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.
