Contrastive Order Learning: A General Framework for Ordinal Regression

Chaewon Lee, Beomjun Shim, Kwang Pyo Choi, Chang-Su Kim
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:65001-65031, 2026.

Abstract

We propose contrastive order learning (ConOrd), a contrastive learning framework for ordinal regression that integrates the strengths of contrastive learning and order learning. While contrastive learning effectively leverages all samples in a batch, it typically ignores the inherent ordering among rank labels. Conversely, order learning explicitly models label ordinality but often relies on local, margin-based comparisons, limiting its ability to capture global ordinal structure. ConOrd addresses these limitations by introducing a contrastive order loss with soft affinity and disparity weights based on rank differences, enabling fine-grained modeling of ordinal relationships across all sample pairs within a batch. Extensive experiments on a range of ordinal regression tasks, including facial age estimation, blind image quality assessment, and blind video quality assessment, demonstrate that ConOrd consistently achieves state-of-the-art performance and generalizes well across diverse ordinal regression scenarios. The source code is available at https://github.com/cwlee00/ConOrd.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-lee26bd, title = {Contrastive Order Learning: A General Framework for Ordinal Regression}, author = {Lee, Chaewon and Shim, Beomjun and Choi, Kwang Pyo and Kim, Chang-Su}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {65001--65031}, 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/lee26bd/lee26bd.pdf}, url = {https://proceedings.mlr.press/v306/lee26bd.html}, abstract = {We propose contrastive order learning (ConOrd), a contrastive learning framework for ordinal regression that integrates the strengths of contrastive learning and order learning. While contrastive learning effectively leverages all samples in a batch, it typically ignores the inherent ordering among rank labels. Conversely, order learning explicitly models label ordinality but often relies on local, margin-based comparisons, limiting its ability to capture global ordinal structure. ConOrd addresses these limitations by introducing a contrastive order loss with soft affinity and disparity weights based on rank differences, enabling fine-grained modeling of ordinal relationships across all sample pairs within a batch. Extensive experiments on a range of ordinal regression tasks, including facial age estimation, blind image quality assessment, and blind video quality assessment, demonstrate that ConOrd consistently achieves state-of-the-art performance and generalizes well across diverse ordinal regression scenarios. The source code is available at https://github.com/cwlee00/ConOrd.} }
Endnote
%0 Conference Paper %T Contrastive Order Learning: A General Framework for Ordinal Regression %A Chaewon Lee %A Beomjun Shim %A Kwang Pyo Choi %A Chang-Su Kim %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-lee26bd %I PMLR %P 65001--65031 %U https://proceedings.mlr.press/v306/lee26bd.html %V 306 %X We propose contrastive order learning (ConOrd), a contrastive learning framework for ordinal regression that integrates the strengths of contrastive learning and order learning. While contrastive learning effectively leverages all samples in a batch, it typically ignores the inherent ordering among rank labels. Conversely, order learning explicitly models label ordinality but often relies on local, margin-based comparisons, limiting its ability to capture global ordinal structure. ConOrd addresses these limitations by introducing a contrastive order loss with soft affinity and disparity weights based on rank differences, enabling fine-grained modeling of ordinal relationships across all sample pairs within a batch. Extensive experiments on a range of ordinal regression tasks, including facial age estimation, blind image quality assessment, and blind video quality assessment, demonstrate that ConOrd consistently achieves state-of-the-art performance and generalizes well across diverse ordinal regression scenarios. The source code is available at https://github.com/cwlee00/ConOrd.
APA
Lee, C., Shim, B., Choi, K.P. & Kim, C.. (2026). Contrastive Order Learning: A General Framework for Ordinal Regression. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:65001-65031 Available from https://proceedings.mlr.press/v306/lee26bd.html.

Related Material