Conformal Margin Risk Minimization: An Envelope Framework for Robust Learning under Label Noise

Yuanjie Shi, Peihong Li, Zijian Zhang, Jana Doppa, Yan Yan
Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, PMLR 300:4276-4284, 2026.

Abstract

Most methods for learning with noisy labels require privileged knowledge such as noise transition matrices, clean subsets or pretrained feature extractors, resources typically unavailable when robustness is most needed. We propose \emph{\textbf{C}onformal \textbf{M}argin \textbf{R}isk \textbf{M}inimization (CMRM)}, a plug-and-play envelope framework that improves \emph{any} classification loss under label noise by adding a single quantile-calibrated regularization term, with no privileged knowledge or training pipeline modification. CMRM measures the confidence margin between the observed label and competing labels, and thresholds it with a conformal quantile estimated per batch to focus training on high-margin samples while suppressing likely mislabeled ones. We derive a learning bound for CMRM under arbitrary label noise requiring only mild regularity of the margin distribution. Across five base methods and six benchmarks with synthetic and real-world noise, CMRM consistently improves accuracy (up to $+3.39$%), reduces conformal prediction set size (up to $-20.44$%) and does not hurt under 0% noise, showing that CMRM captures a method-agnostic uncertainty signal that existing mechanisms did not exploit.

Cite this Paper

BibTeX
@InProceedings{pmlr-v300-shi26b, title = { Conformal Margin Risk Minimization: An Envelope Framework for Robust Learning under Label Noise }, author = {Shi, Yuanjie and Li, Peihong and Zhang, Zijian and Doppa, Jana and Yan, Yan}, booktitle = {Proceedings of The 29th International Conference on Artificial Intelligence and Statistics}, pages = {4276--4284}, year = {2026}, editor = {Khan, Emtiyaz and Li, Yingzhen and Solin, Arno and Ramdas, Aaditya}, volume = {300}, series = {Proceedings of Machine Learning Research}, month = {02--05 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v300/main/assets/shi26b/shi26b.pdf}, url = {https://proceedings.mlr.press/v300/shi26b.html}, abstract = { Most methods for learning with noisy labels require privileged knowledge such as noise transition matrices, clean subsets or pretrained feature extractors, resources typically unavailable when robustness is most needed. We propose \emph{\textbf{C}onformal \textbf{M}argin \textbf{R}isk \textbf{M}inimization (CMRM)}, a plug-and-play envelope framework that improves \emph{any} classification loss under label noise by adding a single quantile-calibrated regularization term, with no privileged knowledge or training pipeline modification. CMRM measures the confidence margin between the observed label and competing labels, and thresholds it with a conformal quantile estimated per batch to focus training on high-margin samples while suppressing likely mislabeled ones. We derive a learning bound for CMRM under arbitrary label noise requiring only mild regularity of the margin distribution. Across five base methods and six benchmarks with synthetic and real-world noise, CMRM consistently improves accuracy (up to $+3.39$%), reduces conformal prediction set size (up to $-20.44$%) and does not hurt under 0% noise, showing that CMRM captures a method-agnostic uncertainty signal that existing mechanisms did not exploit. } }
Endnote
%0 Conference Paper %T Conformal Margin Risk Minimization: An Envelope Framework for Robust Learning under Label Noise %A Yuanjie Shi %A Peihong Li %A Zijian Zhang %A Jana Doppa %A Yan Yan %B Proceedings of The 29th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2026 %E Emtiyaz Khan %E Yingzhen Li %E Arno Solin %E Aaditya Ramdas %F pmlr-v300-shi26b %I PMLR %P 4276--4284 %U https://proceedings.mlr.press/v300/shi26b.html %V 300 %X Most methods for learning with noisy labels require privileged knowledge such as noise transition matrices, clean subsets or pretrained feature extractors, resources typically unavailable when robustness is most needed. We propose \emph{\textbf{C}onformal \textbf{M}argin \textbf{R}isk \textbf{M}inimization (CMRM)}, a plug-and-play envelope framework that improves \emph{any} classification loss under label noise by adding a single quantile-calibrated regularization term, with no privileged knowledge or training pipeline modification. CMRM measures the confidence margin between the observed label and competing labels, and thresholds it with a conformal quantile estimated per batch to focus training on high-margin samples while suppressing likely mislabeled ones. We derive a learning bound for CMRM under arbitrary label noise requiring only mild regularity of the margin distribution. Across five base methods and six benchmarks with synthetic and real-world noise, CMRM consistently improves accuracy (up to $+3.39$%), reduces conformal prediction set size (up to $-20.44$%) and does not hurt under 0% noise, showing that CMRM captures a method-agnostic uncertainty signal that existing mechanisms did not exploit.
APA
Shi, Y., Li, P., Zhang, Z., Doppa, J. & Yan, Y.. (2026). Conformal Margin Risk Minimization: An Envelope Framework for Robust Learning under Label Noise . Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 300:4276-4284 Available from https://proceedings.mlr.press/v300/shi26b.html.

Related Material