Calibration Results for Established Post-hoc Calibration Methods

This archive contains model output files from the application of 20 established post-hoc calibration methods across 15 different calibration tasks. The included methods are: HB, Iso, BBQ, ENIR, TS, VS, MS, Beta, ScaBin, Dir, GP, DIAG, DEC, IMax, SoftBin, Spline, EC, LECE, SCTL, and TS+LECE.

Details of the calibration tasks, experimental setup, and the implementation of these methods can be found in the comparison section of the GitHub repository: https://github.com/WenjianHuang93/h-Calibration

For a comprehensive description of the benchmarking methods and tasks, please refer to the following publication:

Huang W, Cao G, Xia J, Chen J, Wang H, Zhang J. h-calibration: Rethinking Classifier Recalibration with Probabilistic Error-Bounded Objective. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2025.