Saved in:
Bibliographic Details
Main Authors: Huang, Zhiming, Morgenstern, Jamie, Roth, Aaron, Zhang, Claire Jie
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2605.09273
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914584314511360
author Huang, Zhiming
Morgenstern, Jamie
Roth, Aaron
Zhang, Claire Jie
author_facet Huang, Zhiming
Morgenstern, Jamie
Roth, Aaron
Zhang, Claire Jie
contents We study online multicalibration beyond the worst-case. We give a single, efficient algorithm which dynamically interpolates between benign and worst-case sequences by adaptively refining a dyadic grid of prediction values. Its error is controlled by the number of leaves in the refinement tree. Our analysis recovers the known $\widetilde O(T^{2/3})$ worst-case-optimal rate for online multicalibration, while simultaneously automatically adapting to easier instances: in the marginal stochastic setting it obtains a rate of $\widetilde O(\sqrt T)$, and for piecewise-stationary means with $J$ segments its rate is $\widetilde O(\sqrt{JT})$. More generally, the rate depends on a threshold-complexity measure of the predictable mean process relative to the group family. We show that this dependence is tight up to logarithmic factors.
format Preprint
id arxiv_https___arxiv_org_abs_2605_09273
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Instance-Adaptive Online Multicalibration
Huang, Zhiming
Morgenstern, Jamie
Roth, Aaron
Zhang, Claire Jie
Machine Learning
We study online multicalibration beyond the worst-case. We give a single, efficient algorithm which dynamically interpolates between benign and worst-case sequences by adaptively refining a dyadic grid of prediction values. Its error is controlled by the number of leaves in the refinement tree. Our analysis recovers the known $\widetilde O(T^{2/3})$ worst-case-optimal rate for online multicalibration, while simultaneously automatically adapting to easier instances: in the marginal stochastic setting it obtains a rate of $\widetilde O(\sqrt T)$, and for piecewise-stationary means with $J$ segments its rate is $\widetilde O(\sqrt{JT})$. More generally, the rate depends on a threshold-complexity measure of the predictable mean process relative to the group family. We show that this dependence is tight up to logarithmic factors.
title Instance-Adaptive Online Multicalibration
topic Machine Learning
url https://arxiv.org/abs/2605.09273