Research map / Label-efficient and robust learning / Recognition and long-tail
Rethinking Confidence Calibration for Failure Prediction
Fei Zhu, Zhen Cheng, Xu-Yao Zhang, Cheng-Lin Liu
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 |
|---|---|---|---|
| train with mixup | ResNet110 · CIFAR-10 | AUROC: 90.76 -> 86.93 | Hurt |
| train with label smoothing | ResNet110 · CIFAR-10 | AUROC: 90.76 -> 81.89 | Hurt |
| train with label smoothing | ResNet110 · CIFAR-10 | AURC: 9.94 -> 24.50 | Hurt |
| train with mixup | ResNet110 · CIFAR-10 | ECE: 4.14 -> 2.97 | Improved |
| train with focal loss | ResNet110 · CIFAR-10 | ECE: 4.14 -> 1.60 | Improved |
| train with focal loss | ResNet110 · CIFAR-10 | AUROC: 90.76 -> 90.70 | No clear effect |
| add temperature scaling | classifiers · CIFAR/ImageNet | failure prediction | No clear effect |
| add SAM (flat minima) to training | ResNet110 · CIFAR-10 | AUROC: 90.76 -> 93.73 | Improved |
| add SWA (flat minima) to training | ResNet110 · CIFAR-10 | AUROC: 90.76 -> 93.69 | Improved |
| combine SAM and SWA (FMFP) | ResNet110 · CIFAR-10 | AUROC: 90.76 -> 93.76 | Improved |
| combine SAM and SWA (FMFP) | ResNet110 · CIFAR-100 | AUROC: 85.00 -> 87.24 | Improved |
| use FMFP instead of CRL | ResNet110 · CIFAR-10 | AUROC: 92.86 -> 93.76 | Improved |
Closest papers by meaning
- Revisiting Confidence Estimation: Towards Reliable Failure Prediction (TPAMI 2023 · similarity 0.96)
- Pointwise Information Measures as Confidence Estimators in Deep Neural Networks: A Comparative Study (ICML 2025 · similarity 0.86)
- Reassessing How to Compare and Improve the Calibration of Machine Learning Models (ICLR 2025 · similarity 0.86)
- Balancing Two Classifiers via A Simplex ETF Structure for Model Calibration (CVPR 2025 · similarity 0.86)
- A Benchmark Study on Calibration (ICLR 2024 · similarity 0.86)
- SURE: SUrvey REcipes for building reliable and robust deep networks (CVPR 2024 · 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.
