{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,2,13]],"date-time":"2026-02-13T07:22:34Z","timestamp":1770967354445,"version":"3.50.1"},"reference-count":56,"publisher":"Frontiers Media SA","license":[{"start":{"date-parts":[[2026,2,13]],"date-time":"2026-02-13T00:00:00Z","timestamp":1770940800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"content-domain":{"domain":["frontiersin.org"],"crossmark-restriction":true},"short-container-title":["Front. Artif. Intell."],"abstract":"<jats:p>Counterfactual explanations methods provide an important tool in the field of interpretable machine learning. Recent advances in this direction have focused on diffusion models to explain a deep classifier. However, these techniques have predominantly focused on problems in computer vision. In this study, we focus on tabular data typical in finance and the social sciences and propose a novel guided reverse process for categorical features based on an approximation to the Gumbel-softmax distribution. Furthermore, we study the effect of the temperature \u03c4 and derive a theoretical bound between the Gumbel-softmax distribution and our proposed approximated distribution. We perform experiments on several large-scale credit lending and other tabular datasets, assessing their performance in terms of the quantitative measures of interpretability, diversity, instability, and validity. These results indicate that our approach outperforms popular baseline methods, producing robust and realistic counterfactual explanations.<\/jats:p>","DOI":"10.3389\/frai.2026.1743495","type":"journal-article","created":{"date-parts":[[2026,2,13]],"date-time":"2026-02-13T06:37:47Z","timestamp":1770964667000},"update-policy":"https:\/\/doi.org\/10.3389\/crossmark-policy","source":"Crossref","is-referenced-by-count":0,"title":["Tabular diffusion counterfactual explanations"],"prefix":"10.3389","volume":"9","author":[{"given":"Wei","family":"Zhang","sequence":"first","affiliation":[{"name":"Electrical Engineering, Columbia University","place":["New York, NY, United States"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Brian","family":"Barr","sequence":"additional","affiliation":[{"name":"Capital One","place":["New York, NY, United States"]}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"John","family":"Paisley","sequence":"additional","affiliation":[{"name":"Electrical Engineering, Columbia University","place":["New York, NY, United States"]}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"1965","published-online":{"date-parts":[[2026,2,13]]},"reference":[{"key":"B1","article-title":"\u201cNeural additive models: Interpretable machine learning with neural nets,\u201d","volume-title":"Advances in Neural Information Processing Systems","author":"Agarwal","year":"2021"},{"key":"B2","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2006.06848","article-title":"Getting a CLUE: a method for explaining uncertainty estimates","author":"Antor\u00e1n","year":"2020","journal-title":"arXiv"},{"key":"B3","article-title":"\u201cDiffusion visual counterfactual explanations,\u201d","volume-title":"Advances in Neural Information Processing Systems","author":"Augustin","year":"2022"},{"key":"B4","doi-asserted-by":"crossref","DOI":"10.1109\/CVPR52688.2022.01767","article-title":"\u201cBlended diffusion for text-driven editing of natural images,\u201d","volume-title":"IEEE Conference on Computer Vision and Pattern Recognition","author":"Avrahami","year":"2022"},{"key":"B5","article-title":"\u201cNode-GAM: Neural generalized additive model for interpretable deep learning,\u201d","volume-title":"International Conference on Learning Representations","author":"Chang","year":"2022"},{"key":"B6","doi-asserted-by":"crossref","DOI":"10.1007\/978-3-030-58112-1_31","article-title":"\u201cMulti-objective counterfactual explanations,\u201d","volume-title":"International Conference on Parallel Problem Solving from Nature","author":"Dandl","year":"2020"},{"key":"B7","article-title":"\u201cDiffusion models beat GANs on image synthesis,\u201d","volume-title":"Advances in Neural Information Processing Systems","author":"Dhariwal","year":"2021"},{"key":"B8","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2211.15089","article-title":"Continuous diffusion for categorical data","author":"Dieleman","year":"2022","journal-title":"arXiv"},{"key":"B9","doi-asserted-by":"crossref","first-page":"693","DOI":"10.1109\/ICDM59182.2024.00078","article-title":"\u201cQUCE: the minimisation and quantification of path-based uncertainty for generative counterfactual explanations,\u201d","volume-title":"2024 IEEE International Conference on Data Mining (ICDM)","author":"Duell","year":"2024"},{"key":"B10","doi-asserted-by":"publisher","first-page":"196","DOI":"10.1016\/j.inffus.2020.07.001","article-title":"Random forest explainability using counterfactual sets","volume":"63","author":"Fern\u00e1ndez","year":"2020","journal-title":"Inform. Fusion"},{"key":"B11","doi-asserted-by":"publisher","first-page":"68","DOI":"10.3390\/jcp5030068","article-title":"Novel actionable counterfactual explanations for intrusion detection using diffusion models","volume":"5","author":"Galwaduge","year":"2025","journal-title":"J. Cybersecur. Privacy"},{"key":"B12","doi-asserted-by":"publisher","first-page":"139","DOI":"10.1145\/3422622","article-title":"Generative adversarial networks","volume":"63","author":"Goodfellow","year":"2020","journal-title":"Commun. ACM"},{"key":"B13","article-title":"\u201cProtein design with guided discrete diffusion,\u201d","volume-title":"Advances in Neural Information Processing Systems","author":"Gruver","year":"2024"},{"key":"B14","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1805.10820","article-title":"Local rule-based explanations of black box decision systems","author":"Guidotti","year":"2018","journal-title":"arXiv"},{"key":"B15","first-page":"619","article-title":"\u201cRocoursenet: Robust training of a prediction aware recourse model,\u201d","volume-title":"Proceedings of the 32nd ACM International Conference on Information and Knowledge Management","author":"Guo","year":""},{"key":"B16","first-page":"577","article-title":"\u201cCounterNet: End-to-end training of prediction aware counterfactual explanations,\u201d","volume-title":"Proceedings of the 29th ACM SIGKDD Conference on Knowledge Discovery and Data Mining","author":"Guo","year":""},{"key":"B17","article-title":"\u201cDenoising diffusion probabilistic models,\u201d","volume-title":"Advances in Neural Information Processing Systems","author":"Ho","year":"2020"},{"key":"B18","article-title":"\u201cArgmax flows and multinomial diffusion: learning categorical distributions,\u201d","volume-title":"Advances in Neural Information Processing Systems","author":"Hoogeboom","year":"2021"},{"key":"B19","doi-asserted-by":"crossref","DOI":"10.1145\/3306618.3314230","article-title":"\u201cGlobal explanations of neural networks: Mapping the landscape of predictions,\u201d","volume-title":"AAAI\/ACM Conference on AI, Ethics, and Society","author":"Ibrahim","year":"2019"},{"key":"B20","article-title":"\u201cCategorical reparameterization with Gumbel-softmax,\u201d","volume-title":"International Conference on Learning Representations","author":"Jang","year":"2017"},{"key":"B21","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1907.09615","article-title":"Towards realistic individual recourse and actionable explanations in black-box decision making systems","author":"Joshi","year":"2019","journal-title":"arXiv"},{"key":"B22","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1312.6114","article-title":"Auto-encoding variational Bayes","author":"Kingma","year":"2013","journal-title":"arXiv"},{"key":"B23","article-title":"\u201cTabDDPM: Modelling tabular data with diffusion models,\u201d","volume-title":"International Conference on Machine Learning","author":"Kotelnikov","year":"2023"},{"key":"B24","doi-asserted-by":"crossref","DOI":"10.1609\/aaai.v38i20.30229","article-title":"\u201cOn the actionability of outcome prediction,\u201d","volume-title":"AAAI Conference on Artificial Intelligence","author":"Liu","year":"2024"},{"key":"B25","article-title":"\u201cDelayed impact of fair machine learning,\u201d","volume-title":"International Conference on Machine Learning","author":"Liu","year":"2018"},{"key":"B26","article-title":"\u201cA unified approach to interpreting model predictions,\u201d","volume-title":"Advances in Neural Information Processing Systems","author":"Lundberg","year":"2017"},{"key":"B27","doi-asserted-by":"crossref","first-page":"710","DOI":"10.1109\/SaTML59370.2024.00041","article-title":"\u201cNavigating the structured what-if spaces: Counterfactual generation via structured diffusion,\u201d","volume-title":"2024 IEEE Conference on Secure and Trustworthy Machine Learning (SaTML)","author":"Madaan","year":"2024"},{"key":"B28","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.1611.00712","article-title":"The concrete distribution: A continuous relaxation of discrete random variables","author":"Maddison","year":"2016","journal-title":"arXiv"},{"key":"B29","doi-asserted-by":"crossref","DOI":"10.1145\/3351095.3372850","article-title":"\u201cExplaining machine learning classifiers through diverse counterfactual explanations,\u201d","volume-title":"Conference on Fairness, Accountability, and Transparency","author":"Mothilal","year":"2020"},{"key":"B30","doi-asserted-by":"publisher","first-page":"18355","DOI":"10.1109\/TNNLS.2025.3580118","article-title":"Counterfactual explanation through latent adjustment in disentangled space of diffusion model","volume":"36","author":"Na","year":"2025","journal-title":"IEEE Trans. Neural Netw. Learn. Syst"},{"key":"B31","doi-asserted-by":"publisher","first-page":"107501","DOI":"10.1016\/j.patcog.2020.107501","article-title":"Handling incomplete heterogeneous data using vaes","volume":"107","author":"Nazabal","year":"2020","journal-title":"Pattern Recogn"},{"key":"B32","first-page":"1488","article-title":"\u201cCounteRGAN: Generating counterfactuals for real-time recourse and interpretability using residual gans,\u201d","volume-title":"Uncertainty in Artificial Intelligence","author":"Nemirovsky","year":"2022"},{"key":"B33","article-title":"\u201cImproved denoising diffusion probabilistic models,\u201d","volume-title":"International Conference on Machine Learning","author":"Nichol","year":"2021"},{"key":"B34","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2502.09992","article-title":"Large language diffusion models","author":"Nie","year":"2025","journal-title":"arXiv"},{"key":"B35","first-page":"274","author":"Panagiotou","year":"2024"},{"key":"B36","article-title":"\u201cCARLA: a Python library to benchmark algorithmic recourse and counterfactual explanation algorithms,\u201d","volume-title":"Neural Information Processing Systems Track on Datasets and Benchmarks","author":"Pawelczyk","year":"2021"},{"key":"B37","doi-asserted-by":"crossref","DOI":"10.1145\/3366423.3380087","article-title":"\u201cLearning model-agnostic counterfactual explanations for tabular data,\u201d","volume-title":"The Web Conference","author":"Pawelczyk","year":"2020"},{"key":"B38","doi-asserted-by":"crossref","DOI":"10.1145\/3375627.3375850","article-title":"\u201cFACE: feasible and actionable counterfactual explanations,\u201d","volume-title":"AAAI\/ACM Conference on AI, Ethics, and Society","author":"Poyiadzi","year":"2020"},{"key":"B39","article-title":"\u201cNeural basis models for interpretability,\u201d","volume-title":"Advances in Neural Information Processing Systems","author":"Radenovic","year":"2022"},{"key":"B40","doi-asserted-by":"crossref","DOI":"10.1609\/aaai.v37i8.26145","article-title":"\u201cDiffusing Gaussian mixtures for generating categorical data,\u201d","volume-title":"AAAI Conference on Artificial Intelligence","author":"Regol","year":"2023"},{"key":"B41","author":"Ribeiro","year":"2016"},{"key":"B42","first-page":"10684","article-title":"\u201cHigh-resolution image synthesis with latent diffusion models,\u201d","volume-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition","author":"Rombach","year":"2022"},{"key":"B43","doi-asserted-by":"publisher","first-page":"2575","DOI":"10.1109\/TAI.2025.3552057","article-title":"GradCFA: A hybrid gradient-based counterfactual and feature attribution explanation algorithm for local interpretation of neural networks","volume":"6","author":"Sanderson","year":"2025","journal-title":"IEEE Trans. Artif. Intellig"},{"key":"B44","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2412.10193","article-title":"Simple guidance mechanisms for discrete diffusion models","author":"Schiff","year":"2024","journal-title":"arXiv"},{"key":"B45","article-title":"\u201cLearning important features through propagating activation differences,\u201d","volume-title":"International Conference on Machine Learning","author":"Shrikumar","year":"2017"},{"key":"B46","article-title":"\u201cDeep unsupervised learning using nonequilibrium thermodynamics,\u201d","volume-title":"International Conference on Machine Learning","author":"Sohl-Dickstein","year":"2015"},{"key":"B47","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2010.02502","article-title":"Denoising diffusion implicit models","author":"Song","year":"2020","journal-title":"arXiv"},{"key":"B48","doi-asserted-by":"publisher","first-page":"828","DOI":"10.1109\/TEVC.2019.2890858","article-title":"One pixel attack for fooling deep neural networks","volume":"23","author":"Su","year":"2019","journal-title":"IEEE Trans. Evol. Comp"},{"key":"B49","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1109\/IJCNN60899.2024.10651395","article-title":"\u201cModel-based counterfactual explanations incorporating feature space attributes for tabular data,\u201d","volume-title":"2024 International Joint Conference on Neural Networks (IJCNN)","author":"Sumiya","year":"2024"},{"key":"B50","doi-asserted-by":"publisher","DOI":"10.48550\/arXiv.2211.16750","article-title":"Score-based continuous-time discrete diffusion models","author":"Sun","year":"2022","journal-title":"arXiv"},{"key":"B51","article-title":"\u201cAxiomatic attribution for deep networks,\u201d","volume-title":"International Conference on Machine Learning","author":"Sundararajan","year":"2017"},{"key":"B52","first-page":"3763","article-title":"\u201cManifold-aligned counterfactual explanations for neural networks,\u201d","volume-title":"International Conference on Artificial Intelligence and Statistics","author":"Tsiourvas","year":"2024"},{"key":"B53","doi-asserted-by":"crossref","DOI":"10.1007\/978-3-030-86520-7_40","article-title":"\u201cInterpretable counterfactual explanations guided by prototypes,\u201d","volume-title":"Joint European Conference on Machine Learning and Knowledge Discovery in Databases","author":"Van Looveren","year":"2021"},{"key":"B54","doi-asserted-by":"publisher","first-page":"841","DOI":"10.2139\/ssrn.3063289","article-title":"Counterfactual explanations without opening the black box: Automated decisions and the GDPR","volume":"31","author":"Wachter","year":"2017","journal-title":"Harv. J. Law"},{"key":"B55","doi-asserted-by":"crossref","first-page":"36","DOI":"10.1145\/3533271.3561722","article-title":"\u201cAn interpretable deep classifier for counterfactual generation,\u201d","volume-title":"Proceedings of the Third ACM International Conference on AI in Finance","author":"Zhang","year":"2022"},{"key":"B56","doi-asserted-by":"crossref","DOI":"10.1609\/aaai.v38i15.29628","article-title":"\u201cGaussian process neural additive models,\u201d","volume-title":"AAAI Conference on Artificial Intelligence","author":"Zhang","year":"2024"}],"container-title":["Frontiers in Artificial Intelligence"],"original-title":[],"link":[{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2026.1743495\/full","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,2,13]],"date-time":"2026-02-13T06:38:02Z","timestamp":1770964682000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.frontiersin.org\/articles\/10.3389\/frai.2026.1743495\/full"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,2,13]]},"references-count":56,"alternative-id":["10.3389\/frai.2026.1743495"],"URL":"https:\/\/doi.org\/10.3389\/frai.2026.1743495","relation":{},"ISSN":["2624-8212"],"issn-type":[{"value":"2624-8212","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,2,13]]},"article-number":"1743495"}}