Saved in:
| Main Author: | |
|---|---|
| Format: | Preprint |
| Published: |
2026
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2603.14092 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866918388213743616 |
|---|---|
| author | Leznik, Michael |
| author_facet | Leznik, Michael |
| contents | The Expected Calibration Error (ece), the dominant calibration metric in machine learning, compares predicted probabilities against empirical frequencies of binary outcomes. This is appropriate when labels are binary events. However, many modern settings produce labels that are themselves probabilities rather than binary outcomes: a radiologist's stated confidence, a teacher model's soft output in knowledge distillation, a class posterior derived from a generative model, or an annotator agreement fraction. In these settings, ece commits a category error - it discards the probabilistic information in the label by forcing it into a binary comparison. The result is not a noisy approximation that more data will correct. It is a structural misalignment that persists and converges to the wrong answer with increasing precision as sample size grows. We introduce the Soft Mean Expected Calibration Error (smece), a calibration metric for settings where labels are of probabilistic nature. The modification to the ece formula is one line: replace the empirical hard-label fraction in each prediction bin with the mean probability label of the samples in that bin. smece reduces exactly to ece when labels are binary, making it a strict generalisation. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_14092 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Soft Mean Expected Calibration Error (SMECE): A Calibration Metric for Probabilistic Labels Leznik, Michael Machine Learning Methodology The Expected Calibration Error (ece), the dominant calibration metric in machine learning, compares predicted probabilities against empirical frequencies of binary outcomes. This is appropriate when labels are binary events. However, many modern settings produce labels that are themselves probabilities rather than binary outcomes: a radiologist's stated confidence, a teacher model's soft output in knowledge distillation, a class posterior derived from a generative model, or an annotator agreement fraction. In these settings, ece commits a category error - it discards the probabilistic information in the label by forcing it into a binary comparison. The result is not a noisy approximation that more data will correct. It is a structural misalignment that persists and converges to the wrong answer with increasing precision as sample size grows. We introduce the Soft Mean Expected Calibration Error (smece), a calibration metric for settings where labels are of probabilistic nature. The modification to the ece formula is one line: replace the empirical hard-label fraction in each prediction bin with the mean probability label of the samples in that bin. smece reduces exactly to ece when labels are binary, making it a strict generalisation. |
| title | Soft Mean Expected Calibration Error (SMECE): A Calibration Metric for Probabilistic Labels |
| topic | Machine Learning Methodology |
| url | https://arxiv.org/abs/2603.14092 |