Research map / Label-efficient and robust learning / Recognition and long-tail
Beyond temperature scaling: Obtaining well-calibrated multi-class probabilities with Dirichlet calibration
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 |
|---|---|---|---|
| Use Dirichlet calibration instead of temperature scaling | Deep neural network classifiers · CIFAR-10 | cw-ECE: 0.0444 to 0.0292 | Helped |
| Use Dirichlet calibration instead of temperature scaling | Deep neural network classifiers · CIFAR-100 | cw-ECE: 0.2271 to 0.2403; 0.1864 to 0.1802 | Mixed |
| Use Dirichlet calibration instead of temperature scaling | Deep neural network classifiers · SVHN | cw-ECE: 0.0384 to 0.0262; 0.0181 to 0.0203 | Mixed |
| Apply Dirichlet calibration with ODIR rather than temperature scaling | Deep neural network classifiers · CIFAR-10 | log-loss: 0.1951 to 0.1822 | Helped |
| Use matrix scaling with ODIR rather than Dirichlet calibration | Deep neural network classifiers · Deep neural network experiments | log-loss: MS-ODIR outperforms Dir-ODIR in 11 out of 14 cases | Mixed |
| Remove off-diagonal coefficients from MS-ODIR (ablation) | MS-ODIR · Deep neural network experiments | log-loss: Almost identical in 6 out of 14 cases; significantly worse in the remaining 8 out of 14 | Mixed |
| Use ODIR rather than L2 regularisation (ablation) | Dirichlet calibration · CIFAR-100 | cw-ECE: 0.4025 to 0.2403 | Helped |
Closest papers by meaning
- On Temperature Scaling and Conformal Prediction of Deep Classifiers (ICML 2025 · similarity 0.89)
- Improving Multi-Class Calibration through Normalization-Aware Isotonic Techniques (ICML 2025 · similarity 0.88)
- Divide et Calibra: Multiclass Local Calibration via Vector Quantization (NeurIPS 2026 · similarity 0.88)
- T-CIL: Temperature Scaling using Adversarial Perturbation for Calibration in Class-Incremental Learning (CVPR 2025 · similarity 0.88)
- On the Limitations of Temperature Scaling for Distributions with Overlaps (ICLR 2024 · similarity 0.88)
- Confidence Calibration of Classifiers with Many Classes (NeurIPS 2024 · 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.