mlsanityData to 14 Oct 2026

MCGrad: Multicalibration at Web Scale

arXiv 2509.19884 ↗First posted 24 Sep 2025cs.LGcited by 5

Not on the map. Why?

Abstract (arXiv)

We propose MCGrad, a novel and scalable multicalibration algorithm. Multicalibration - calibration in subgroups of the data - is an important property for the performance of machine learning-based systems. Existing multicalibration methods have thus far received limited traction in industry. We argue that this is because existing methods (1) require such subgroups to be manually specified, which ML practitioners often struggle with, (2) are not scalable, or (3) may harm other notions of model performance such as log loss and Area Under the Precision-Recall Curve (PRAUC). MCGrad does not require explicit specification of protected groups, is scalable, and often improves other ML evaluation metrics instead of harming them. MCGrad has been in production at Meta, and is now part of hundreds of production models. We present results from these deployments as well as results on public datasets. We provide an open source implementation of MCGrad at https://github.com/facebookincubator/MCGrad.

Key facts. Cited by 5 papers in the map (data through 14 Oct 2026). Builds on 31 of its 73 references that are in the map. First posted on arXiv 24 Sep 2025.

Builds on

31 of 73 references.

References · 31

  1. 2016XGBoost: A Scalable Tree Boosting Systemcited by 5,166
  2. 2017On Calibration of Modern Neural Networkscited by 4,960
  3. 2017What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?cited by 2,352
  4. 2021Retiring Adult: New Datasets for Fair Machine Learningcited by 449
  5. 2018Improving fairness in machine learning systems: What do industry practitioners need?cited by 262
  6. 2020Mitigating Bias in Calibration Error Estimationcited by 101
  7. 2021Calibrating Predictions to Decisions: A Novel Approach to Multi-Class Calibrationcited by 79
  8. 2021Classifier Calibration: A survey on how to assess and improve predicted class probabilitiescited by 73
  9. 2020Moment Multicalibration for Uncertainty Estimationcited by 63
  10. 2023Smooth ECE: Principled Reliability Diagrams via Kernel Smoothingcited by 63
  11. 2024Multicalibration for Confidence Scoring in LLMscited by 46
  12. 2022Metrics of calibration for probabilistic predictionscited by 32
19 more
  1. 2023Swap Agnostic Learning, or Characterizing Omniprediction via Multicalibrationcited by 31
  2. 2023A Unifying Perspective on Multi-Calibration: Game Dynamics for Multi-Objective Learningcited by 30
  3. 2020Sample Complexity of Uniform Convergence for Multicalibrationcited by 24
  4. 2021On the Calibration and Uncertainty of Neural Learning to Rank Modelscited by 24
  5. 2024On Calibration of LLM-based Guard Models for Reliable Content Moderationcited by 23
  6. 2023Loss Minimization Yields Multicalibration for Large Neural Networkscited by 19
  7. 2021Measuring and Improving Model-Moderator Collaboration using Uncertainty Estimationcited by 18
  8. 2024When is Multicalibration Post-Processing Necessary?cited by 18
  9. 2020FaiR-N: Fair and Robust Neural Networks for Structured Datacited by 16
  10. 2021Obtaining Calibrated Probabilities with Personalized Ranking Modelscited by 15
  11. 2024Bridging Multicalibration and Out-of-distribution Generalization Beyond Covariate Shiftcited by 10
  12. 2018Randomized Gradient Boosting Machinecited by 9
  13. 2021Looper: An end-to-end ML platform for product decisionscited by 9
  14. 2025Discretization-free Multicalibration through Loss Minimization over Tree Ensemblescited by 7
  15. 2022Calibration of P-values for calibration and for deviation of a subpopulation from the full populationcited by 4
  16. 2024Deep Ensemble Shape Calibration: Multi-Field Post-hoc Calibration in Online Advertisingcited by 4
  17. 2025Measuring multi-calibrationcited by 4
  18. 2025Calibrated Recommendations: Survey and Future Directionscited by 3
  19. 2022Calibration Matters: Tackling Maximization Bias in Large-scale Advertising Recommendation Systemscited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 5.

5 papers.

  1. 2026On the Convergence of Multicalibration Gradient Boostingcited by 1
  2. 2026Divide et Calibra: Multiclass Local Calibration via Vector Quantizationcited by 1
  3. 2026Multicalibration for Unbiased Model-Based Prevalence Estimationcited by 0
  4. 2026Team DACTYL at PAN 2026: Bayesian Data Mixing and Empirical X-risk Minimization for AI-text Detectioncited by 0
  5. 2026A CaLLM Judge is a Better Judge: Mitigating Bias in LLM-as-a-judgecited by 0