Calibrating Generative Models to Distributional Constraints

Henry Smith, Nathaniel Lee Diamant, Brian L. Trippe
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:113976-114023, 2026.

Abstract

Generative models frequently suffer miscalibration, wherein statistics of the sampling distribution—such as the fraction of generations in a given class—deviate from desired values. We frame calibration as a constrained optimization problem and seek the closest model in Kullback-Leibler divergence satisfying a calibration constraint. To address the intractability of imposing these constraints exactly, we introduce two surrogate objectives for fine-tuning: (1) the relax loss, which replaces the constraint with a miscalibration penalty, and (2) the reward loss, which converts calibration into a reward fine-tuning problem. We demonstrate that these approaches substantially reduce calibration error across hundreds of simultaneous constraints and models with up to nine billion parameters, spanning applications in protein design, image generation, and language modeling. Code is available at https://github.com/smithhenryd/cgm.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-smith26a, title = {Calibrating Generative Models to Distributional Constraints}, author = {Smith, Henry and Diamant, Nathaniel Lee and Trippe, Brian L.}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {113976--114023}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/smith26a/smith26a.pdf}, url = {https://proceedings.mlr.press/v306/smith26a.html}, abstract = {Generative models frequently suffer miscalibration, wherein statistics of the sampling distribution—such as the fraction of generations in a given class—deviate from desired values. We frame calibration as a constrained optimization problem and seek the closest model in Kullback-Leibler divergence satisfying a calibration constraint. To address the intractability of imposing these constraints exactly, we introduce two surrogate objectives for fine-tuning: (1) the relax loss, which replaces the constraint with a miscalibration penalty, and (2) the reward loss, which converts calibration into a reward fine-tuning problem. We demonstrate that these approaches substantially reduce calibration error across hundreds of simultaneous constraints and models with up to nine billion parameters, spanning applications in protein design, image generation, and language modeling. Code is available at https://github.com/smithhenryd/cgm.} }
Endnote
%0 Conference Paper %T Calibrating Generative Models to Distributional Constraints %A Henry Smith %A Nathaniel Lee Diamant %A Brian L. Trippe %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-smith26a %I PMLR %P 113976--114023 %U https://proceedings.mlr.press/v306/smith26a.html %V 306 %X Generative models frequently suffer miscalibration, wherein statistics of the sampling distribution—such as the fraction of generations in a given class—deviate from desired values. We frame calibration as a constrained optimization problem and seek the closest model in Kullback-Leibler divergence satisfying a calibration constraint. To address the intractability of imposing these constraints exactly, we introduce two surrogate objectives for fine-tuning: (1) the relax loss, which replaces the constraint with a miscalibration penalty, and (2) the reward loss, which converts calibration into a reward fine-tuning problem. We demonstrate that these approaches substantially reduce calibration error across hundreds of simultaneous constraints and models with up to nine billion parameters, spanning applications in protein design, image generation, and language modeling. Code is available at https://github.com/smithhenryd/cgm.
APA
Smith, H., Diamant, N.L. & Trippe, B.L.. (2026). Calibrating Generative Models to Distributional Constraints. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:113976-114023 Available from https://proceedings.mlr.press/v306/smith26a.html.

Related Material