Search papers, labs, and topics across Lattice.
The authors introduce CalArena, a large-scale benchmark for post-hoc calibration, encompassing nearly 2000 experiments across tabular and computer vision tasks, using predictions from diverse models. They advocate for Post-Hoc Improvement (PHI) in proper scoring rules as a more principled evaluation metric compared to traditional calibration error estimators. Their comprehensive empirical study reveals that smooth calibration functions outperform binning-based approaches, dedicated multiclass methods are crucial in high-dimensional settings, and generic machine learning models require calibration-specific designs.
Stop relying on small-scale, inconsistent evaluations: CalArena offers a large-scale, standardized benchmark revealing that smooth calibration functions and dedicated multiclass methods are essential for effective post-hoc calibration.
Reliable probability estimates are critical in many machine learning applications, yet modern classifiers are often poorly calibrated. Post-hoc calibration provides a simple and widely used solution, but the large number of proposed methods, combined with small-scale and inconsistent evaluations, makes it difficult to determine which approaches are truly effective in practice. We introduce a large-scale, standardized benchmark for post-hoc calibration, covering nearly 2000 experiments across tabular and computer vision tasks, including binary, multiclass, and large-scale classification settings. Our benchmark aggregates predictions from a diverse set of classical models, modern deep learning architectures, and foundation models, and provides unified, reproducible implementations of dozens of calibration methods within a common evaluation framework. We argue that Post-Hoc Improvement (PHI) in proper scoring rules offers a principled alternative to traditional calibration error estimators for comparing post-hoc methods, capturing both calibration quality and potential degradation to the model's predictive performance. Using this framework, we conduct the most comprehensive empirical study of post-hoc calibration to date. Our results reveal consistent patterns across domains: smooth calibration functions outperform binning-based approaches, dedicated multiclass methods are essential in high-dimensional settings, and generic machine learning models are not competitive without calibration-specific design. To facilitate future research, we release all data, code, and evaluation tools, providing a plug-and-play benchmark for developing and comparing calibration methods.
