Research map / Label-efficient and robust learning / OOD and anomaly detection
ViM: Out-of-Distribution With Virtual-Logit Matching
Check 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 |
|---|---|---|---|
| Combine feature residual information with class-dependent logits using a scaled virtual OOD logit | ImageNet-1K OOD detection with BiT-S-R101×1 · Four OOD benchmark average | AUROC: 90.91% vs 86.62% | Helped |
| Combine feature residual information with class-dependent logits using a scaled virtual OOD logit | ImageNet-1K OOD detection with ViT-B/16 · Four OOD benchmark average | AUROC: 96.23 vs 96.02 | Helped |
| Use the combined feature-and-logit score rather than the strongest listed baseline | BiT-S-R101×1 · iNaturalist | AUROC: 92.95 -> 89.30 | Hurt |
| Add taxonomy grouping to ViM (ablation) | BiT-S-R101×1 · OpenImage-O | AUROC: 91.54 -> 91.92 | Helped |
| Add taxonomy grouping to ViM (ablation) | BiT-S-R101×1 · iNaturalist | AUROC: 89.30 -> 90.16 | Helped |
| Add taxonomy grouping to ViM (ablation) | BiT-S-R101×1 · ImageNet-O | AUROC: 83.87 -> 83.43 | Hurt |
| Add taxonomy grouping to ViM (ablation) | ViT-B/16 · Texture | AUROC: 95.34 -> 95.29 | No clear effect |
| Vary the principal-space dimension D across a wide range (ablation) | ViM · ImageNet-1K OOD benchmarks | AUROC | No clear effect |
| Perturb the matching parameter α away from its training-set scale match (ablation) | ViM · ImageNet-1K OOD benchmarks | AUROC | Hurt |
Closest papers by meaning
- Revisiting Logit Distributions for Reliable Out-of-Distribution Detection (NeurIPS 2025 · similarity 0.91)
- Image Background Serves as Good Proxy for Out-of-distribution Data (ICLR 2024 · similarity 0.90)
- A Systematic Analysis of Out-of-Distribution Detection Under Representation and Training Paradigm Shifts (NeurIPS 2026 · similarity 0.89)
- MultiOOD: Scaling Out-of-Distribution Detection for Multiple Modalities (NeurIPS 2024 · similarity 0.89)
- Is Complex Training Necessary for Long-Tailed OOD Detection? A Re-think from Feature Geometry (NeurIPS 2026 · similarity 0.89)
- ImageNet-OOD: Deciphering Modern Out-of-Distribution Detection Algorithms (ICLR 2024 · similarity 0.89)
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.