Research map / Label-efficient and robust learning / Recognition and long-tail
What Can we Learn From The Selective Prediction And Uncertainty Estimation Performance Of 523 Imagenet Classifiers?
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 |
|---|---|---|---|
| Add knowledge distillation during training (ablation) | ImageNet classifiers · ImageNet | AUROC | Helped |
| Add knowledge distillation during training (ablation) | ImageNet classifiers · ImageNet | ECE | Helped |
| Use temperature scaling (ablation) | ImageNet classifiers · ImageNet | AUROC | Mixed |
| Use temperature scaling (ablation) | ImageNet classifiers · ImageNet | selective performance | Helped |
| Use temperature scaling (ablation) | ImageNet classifiers · ImageNet | ECE | Mixed |
| Pretrain on ImageNet-21k without other additions (ablation) | otherwise identical classifiers · ImageNet | ECE | Helped |
| Pretrain on ImageNet-21k without other additions (ablation) | otherwise identical classifiers · ImageNet | AUROC | Hurt |
| Replace softmax confidence with MC-dropout predictive entropy (ablation) | the same classifiers · ImageNet | AUROC | Hurt |
| Replace softmax confidence with MC-dropout predictive entropy (ablation) | the same classifiers · ImageNet | selective performance | Hurt |
| Use ViT architecture under the successful training regimes studied | ImageNet classifiers · ImageNet | AUROC: 88.5 vs 87.7 | Helped |
| Use ViT architecture under the successful training regimes studied | ImageNet classifiers · ImageNet | coverage at 99% selective accuracy: beyond 30% | Helped |
| Add a linear probe to zero-shot CLIP (ablation) | CLIP ViT-B/32 · ImageNet | ECE: 1.3% | Mixed |
Closest papers by meaning
- SURE: SUrvey REcipes for building reliable and robust deep networks (CVPR 2024 · similarity 0.89)
- Simple and Scalable Predictive Uncertainty Estimation using Deep Ensembles (NeurIPS 2017 · similarity 0.88)
- Hierarchical Selective Classification (NeurIPS 2024 · similarity 0.88)
- Training-Free Uncertainty Estimation for Dense Regression: Sensitivity as a Surrogate (AAAI 2022 · similarity 0.87)
- What Does It Take to Build a Performant Selective Classifier? (NeurIPS 2025 · similarity 0.86)
- Deep Probabilistic Supervision for Image Classification (NeurIPS 2026 · 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.