Research map / Label-efficient and robust learning / OOD and anomaly detection
Decoupling MaxLogit 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 |
|---|---|---|---|
| decouple MaxCosine from MaxNorm and combine them with a tunable weight | MaxLogit OOD scoring · CIFAR-100 | AUROC: 80.96 -> 82.34 | Helped |
| combine MaxCosine from a Focal-loss model with MaxNorm from a Center-loss model | DML · CIFAR-100 (six OOD datasets, average) | AUROC: 85.73 -> 91.57 | Helped |
| combine MaxCosine from a Focal-loss model with MaxNorm from a Center-loss model | DML · CIFAR-10 (six OOD datasets, average) | AUROC: 94.98 -> 98.00 | Helped |
| combine MaxCosine from a Focal-loss model with MaxNorm from a Center-loss model | DML · ImageNet (four OOD datasets, average) | AUROC: 90.55 -> 93.16 | Helped |
| use MaxCosine alone instead of combining MaxCosine and MaxNorm (ablation) | DML on CIFAR-100 · CIFAR-100 | AUROC: 81.78 -> 82.34 | Hurt |
| use MaxNorm alone instead of combining MaxCosine and MaxNorm (ablation) | DML on CIFAR-100 · CIFAR-100 | AUROC: 56.93 -> 82.34 | Hurt |
| replace the linear classifier with a cosine classifier (ablation) | logit-based scoring methods · CIFAR-100 | AUROC: 82.34 -> 85.34 | Helped |
| replace the linear classifier with a cosine classifier (ablation) | MaxNorm · CIFAR-100 | AUROC: 56.93 -> 76.39 | Helped |
| train the cosine-classifier model with Center loss (ablation) | DML · CIFAR-100 | AUROC: 85.34 -> 89.86 | Helped |
| train the cosine-classifier model with Focal loss (ablation) | MaxCosine · CIFAR-100 | AUROC: 77.40 -> 90.90 | Helped |
| normalize and uniformly couple the MCF and MNC scores (ablation) | DML+ · Textures | AUROC: 87.89 -> 89.70 | Helped |
| use MaxCosine from the Focal-loss model rather than MaxNorm from the Center-loss model (ablation) | DML+ components · CIFAR-100 Textures | AUROC: 85.52 -> 91.74 | Helped |
Closest papers by meaning
- Revisiting Logit Distributions for Reliable Out-of-Distribution Detection (NeurIPS 2025 · similarity 0.91)
- ConjNorm: Tractable Density Estimation for Out-of-Distribution Detection (ICLR 2024 · similarity 0.88)
- Double Descent Meets Out-of-Distribution Detection: Theoretical Insights and Empirical Analysis on the Role of Model Complexity (NeurIPS 2025 · similarity 0.87)
- EAT: Towards Long-Tailed Out-of-Distribution Detection (AAAI 2024 · similarity 0.86)
- Long-Tailed Out-of-Distribution Detection: Prioritizing Attention to Tail (AAAI 2025 · similarity 0.86)
- Beyond Clean Training Data: A Versatile and Model-Agnostic Framework for Out-of-Distribution Detection with Contaminated Training Data (CVPR 2025 · similarity 0.86)
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.