mlsanityData to 7 Oct 2026

Calibrating Generative Models to Distributional Constraints

arXiv 2510.10020 ↗First posted 11 Oct 2025stat.ML cs.LG q-bio.BMcited by 5

Not on the map. Why?

Abstract (arXiv)

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.

Key facts. Cited by 5 papers in the map (data through 7 Oct 2026). Builds on 41 of its 96 references that are in the map. First posted on arXiv 11 Oct 2025.

Builds on

41 of 96 references.

References · 41

  1. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 29,915
  2. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  3. 2020Score-Based Generative Modeling through Stochastic Differential Equationscited by 8,964
  4. 2021Diffusion Models Beat GANs on Image Synthesiscited by 8,185
  5. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  6. 2022Classifier-Free Diffusion Guidancecited by 5,717
  7. 2017On Calibration of Modern Neural Networkscited by 4,860
  8. 2017Deep reinforcement learning from human preferencescited by 3,846
  9. 2021Convex optimizationcited by 2,479
  10. 2015Variational Inference with Normalizing Flowscited by 2,161
  11. 2024Gemma 2: Improving Open Language Models at a Practical Sizecited by 1,928
  12. 2019StarGAN v2: Diverse Image Synthesis for Multiple Domainscited by 1,095
29 more
  1. 2023Training Diffusion Models with Reinforcement Learningcited by 939
  2. 2023Diffusion Model Alignment Using Direct Preference Optimizationcited by 767
  3. 2023Bias and Fairness in Large Language Models: A Surveycited by 626
  4. 2023DPOK: Reinforcement Learning for Fine-tuning Text-to-Image Diffusion Modelscited by 431
  5. 2023TinyStories: How Small Can Language Models Be and Still Speak Coherent English?cited by 425
  6. 2013Black Box Variational Inferencecited by 317
  7. 2019Monte Carlo Gradient Estimation in Machine Learningcited by 281
  8. 2022Diffusion probabilistic modeling of protein backbones in 3D for the motif-scaffolding problemcited by 229
  9. 2024Adjoint Matching: Fine-tuning Flow and Diffusion Generative Models with Memoryless Stochastic Optimal Controlcited by 222
  10. 2023SE(3) diffusion model with application to protein backbone generationcited by 214
  11. 2019Evaluating model calibration in classificationcited by 165
  12. 2023Fair Diffusion: Instructing Text-to-Image Generation Models on Fairnesscited by 149
  13. 2024Fine-Tuning of Continuous-Time Diffusion Models as Entropy-Regularized Controlcited by 136
  14. 2023Aligning Language Models with Preferences through f-divergence Minimizationcited by 118
  15. 2020A Distributional Approach to Controlled Text Generationcited by 117
  16. 2023Finetuning Text-to-Image Diffusion Models for Fairnesscited by 90
  17. 2024Normalizing Flows are Capable Generative Modelscited by 89
  18. 2018Scalable End-to-End Autonomous Vehicle Testing via Rare-event Simulationcited by 76
  19. 2023Class-Balancing Diffusion Modelscited by 47
  20. 2024DEFT: Efficient Fine-Tuning of Diffusion Models by Learning the Generalised h-transformcited by 44
  21. 2024Fine-tuning of diffusion models via stochastic control: entropy regularization and beyondcited by 42
  22. 2019Enforcing Statistical Constraints in Generative Adversarial Networks for Modeling Chaotic Dynamical Systemscited by 33
  23. 2024Out of Many, One: Designing and Scaffolding Proteins at the Scale of the Structural Universe with Genie 2cited by 33
  24. 2017Do GANs actually learn the distribution? An empirical studycited by 23
  25. 2024Constrained Diffusion Models via Dual Trainingcited by 19
  26. 2025Constrained Discrete Diffusioncited by 16
  27. 2024PROUD: PaRetO-gUided Diffusion Model for Multi-objective Generationcited by 15
  28. 2025Composition and Alignment of Diffusion Models using Constrained Learningcited by 8
  29. 2025Minimum-Excess-Work Guidancecited by 1

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 5.

5 papers.

  1. 2026Bridging the Simulation-to-Experiment Gap with Generative Models using Adversarial Distribution Alignmentcited by 1
  2. 2026The Reward Was in Your Data All Along: Correcting Flow Matching with Discriminator-Guided RLcited by 0
  3. 2026Calibrating Generative Models to Feature Distributions with MMD Finetuningcited by 0
  4. 2026A Mean-Field Framework for Inference-Time Distributional Control of Diffusion Modelscited by 0
  5. 2026SupraTITO: Transferable Generative Molecular Dynamics for Supramolecular Systemscited by 0