Overview

For classifiers and language models, calibration asks whether confidence matches empirical correctness and develops post-hoc methods such as temperature scaling, isotonic regression, and local corrections, alongside training-time objectives. Online and multicalibration methods extend reliability to changing data, subgroups, and selective prediction, while conformal methods provide coverage guarantees for prediction sets or intervals. A related line calibrates mechanistic models and simulators to measurements using Bayesian inference, evolutionary optimization, differentiable surrogates, and data assimilation in areas such as climate, traffic, medicine, and engineering. Across settings from CIFAR and ImageNet to language-model question answering and clinical data, a central challenge is that aggregate calibration may fail for particular groups or under distribution shift, while flexible corrections require representative validation data and may affect accuracy or decisions.

Latest papers

245 papers40 papers in the last four weeks, up 8% on the four weeks before. 0.3% of all new papers.

Jun 22, 2025

Improving Mixup Calibration with Wasserstein Distributionally Robust Optimization

In many real-world applications, ensuring the robustness and stability of deep neural networks (DNNs) is crucial, particularly for image classification tasks that encounter various input perturbations. While Mixup-based data augmentation techniques have been widely adopted to enhance the resilience of trained models against such perturbations, our experiments reveal an important corruption robustness-calibration trade-off: stronger Mixup-based augmentation can improve robustness against corrupted data while substantially increasing expected calibration error (ECE). To address this challenge, we introduce DRO-Augment, a framework that integrates Wasserstein Distributionally Robust Optimization (W-DRO) with various Mixup-based data augmentation strategies to mitigate this trade-off. Our method substantially reduces ECE under strong Mixup-based augmentation while largely preserving corruption accuracy across CIFAR-10, CIFAR-100, CIFAR-10-C, and CIFAR-100-C. On the theoretical side, we establish novel generalization error bounds for neural networks trained using a variation-regularized loss function with augmented data, closely related to the W-DRO problem. Furthermore, we introduce a refined CIFAR-C benchmark that corrects inconsistencies in corruption intensities, providing a more reliable evaluation for future robustness research.
May 29, 2025

Towards Understanding The Calibration Benefits of Sharpness-Aware Minimization

Deep neural networks have been increasingly used in safety-critical applications such as medical diagnosis and autonomous driving. However, many studies suggest that they are prone to being poorly calibrated and have a propensity for overconfidence, which may have disastrous consequences. In this paper, unlike standard training such as stochastic gradient descent, we show that the recently proposed sharpness-aware minimization (SAM) counteracts this tendency towards overconfidence. The theoretical analysis suggests that SAM allows us to learn models that are already well-calibrated by implicitly maximizing the entropy of the predictive distribution. Inspired by this finding, we further propose a variant of SAM, coined as CSAM, to ameliorate model calibration. Extensive experiments on various datasets, including ImageNet-1K, demonstrate the benefits of SAM in reducing calibration error. Meanwhile, CSAM performs even better than SAM and consistently achieves lower calibration error than other approaches
May 28, 2025

Optimizable Model Averaging for Ensembling Heterogeneous Model Outputs

Foundation model ensembles can improve accuracy and robustness, but full fine-tuning and parameter-space merging are often expensive or impossible when candidate models have different architectures. We propose Optimizable Model Averaging (OMA), a black-box approach that learns nonnegative model-combination weights, one scalar coefficient per model, to average the candidate models' predictive probabilities, leaving all model parameters frozen, by minimizing regularized predictive entropy. A central challenge is that minimizing predictive entropy can favour confidently incorrect models: increased confidence need not imply improved accuracy or calibration. We therefore examine regularization as a mechanism for controlling adaptation and empirically evaluate its effect on calibration, observing lower expected calibration error in selected configurations, alongside better performance-calibration trade-offs than model output averaging. As foundation models continue to advance, these approaches will enable the incorporation of future, potentially much better models to improve performance on challenging text and image classification tasks.
Nov 20, 2022

MetaMax: Improved Open-Set Deep Neural Networks via Weibull Calibration

Open-set recognition refers to the problem in which classes that were not seen during training appear at inference time. This requires the ability to identify instances of novel classes while maintaining discriminative capability for closed-set classification. OpenMax was the first deep neural network-based approach to address open-set recognition by calibrating the predictive scores of a standard closed-set classification network. In this paper we present MetaMax, a more effective post-processing technique that improves upon contemporary methods by directly modeling class activation vectors. MetaMax removes the need for computing class mean activation vectors (MAVs) and distances between a query image and a class MAV as required in OpenMax. Experimental results show that MetaMax outperforms OpenMax and is comparable in performance to other state-of-the-art approaches.
Jun 24, 2022

How many labelers do you have? A closer look at gold-standard labels

The construction of most supervised learning datasets revolves around collecting multiple labels for each instance, then aggregating the labels to form a type of "true" label. We question the wisdom of this pipeline by developing a (stylized) theoretical model of this process and analyzing its statistical consequences, showing how access to non-aggregated label information can make training well-calibrated models more feasible than it is with cleaned labels. The entire story, however, is subtle, and the contrasts between aggregated and fuller label information depend on the particulars of the problem, where estimators that use aggregated information exhibit robust but slower rates of convergence, while estimators that can effectively leverage all labels converge more quickly if they have fidelity to (or can learn) the true labeling process. The theory makes several predictions for real-world datasets, including when non-aggregate labels should improve learning performance, which we test to corroborate the validity of our predictions.