{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,7]],"date-time":"2026-07-07T21:22:02Z","timestamp":1783459322225,"version":"3.55.0"},"reference-count":40,"publisher":"Wiley","issue":"2","license":[{"start":{"date-parts":[[2026,4,9]],"date-time":"2026-04-09T00:00:00Z","timestamp":1775692800000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"},{"start":{"date-parts":[[2026,4,9]],"date-time":"2026-04-09T00:00:00Z","timestamp":1775692800000},"content-version":"tdm","delay-in-days":0,"URL":"http:\/\/doi.wiley.com\/10.1002\/tdm_license_1.1"}],"content-domain":{"domain":["onlinelibrary.wiley.com"],"crossmark-restriction":true},"short-container-title":["AI Magazine"],"published-print":{"date-parts":[[2026,6]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>\n                    AI systems are rapidly transitioning from laboratory demonstrations to decision\u2010making technologies deployed in high\u2010stakes domains. Yet reliability remains a primary obstacle to responsible adoption: discriminative models can be confidently wrong under out\u2010of\u2010distribution (OOD) inputs, and foundation models (FMs) such as large language models (LLMs) can generate fluent but untruthful, harmful, or misaligned outputs. My research develops the foundations of\n                    <jats:italic>reliable machine learning with minimal human supervision<\/jats:italic>\n                    , unifying algorithms, and theory that make reliability a first\u2010class objective alongside accuracy. I advance unknown\u2010aware learning through automated outlier generation, introducing feature\u2010 and input\u2010space synthesis frameworks that regularize decision boundaries and improve interpretability. I further establish principled methods for learning \u201cin the wild\u201d by leveraging unlabeled deployment data under mixture and contamination models, with theoretical guarantees and state\u2010of\u2010the\u2010art performance for OOD detection and generalization under diverse shifts. Finally, I design reliability frameworks for FMs by exploiting unlabeled signals to detect hallucinations, defend against malicious prompts in vision\u2013language models, and denoise noisy preference data for more dependable alignment. Collectively, these contributions provide a cohesive toolkit for deploying AI systems that remain accurate, calibrated, and trustworthy in open\u2010world\u00a0environments.\n                  <\/jats:p>","DOI":"10.1002\/aaai.70058","type":"journal-article","created":{"date-parts":[[2026,4,10]],"date-time":"2026-04-10T06:04:51Z","timestamp":1775801091000},"update-policy":"https:\/\/doi.org\/10.1002\/crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Teach AI What It Doesn't Know"],"prefix":"10.1002","volume":"47","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-8080-7861","authenticated-orcid":false,"given":"Sean","family":"Du","sequence":"first","affiliation":[{"name":"College of Computing and Data Science Nanyang Technological University Singapore Singapore"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"311","published-online":{"date-parts":[[2026,4,9]]},"reference":[{"key":"e_1_2_8_2_1","unstructured":"Bai H. G.Canal X.Du J.Kwon R. D.Nowak andY.Li.2023. \u201cFeed Two Birds with One Scone: Exploiting Wild Data for Both Out\u2010of\u2010Distribution Generalization and Detection.\u201d InInternational Conference on Machine Learning."},{"key":"e_1_2_8_3_1","unstructured":"Bai H. X.Du K.Rainey S.Parameswaran andY.Li.2025. \u201cOut\u2010of\u2010Distribution Learning with Human Feedback.\u201dTransactions on Machine Learning Research."},{"key":"e_1_2_8_4_1","unstructured":"Constantinou C. G.Ioannides A.Chadha A.Elkins andE.Simpson. \u201cOut\u2010of\u2010Distribution Detection with Attention Head Masking for Multimodal Document Classification.\u201dNature Scientific Reports."},{"key":"e_1_2_8_5_1","unstructured":"C. R. Association.2019. \u201cArtificial Intelligence Roadmap.\u201dhttps:\/\/cra.org\/ccc\/ai\u2010roadmap\u2010introduction\/."},{"key":"e_1_2_8_6_1","unstructured":"Du X. Z.Fang I.Diakonikolas andY.Li.2024. \u201cHow Does Unlabeled Data Provably Help Out\u2010of\u2010Distribution Detection?\u201d InInternational Conference on Learning Representations."},{"key":"e_1_2_8_7_1","unstructured":"Du X. R.Ghosh R.Sim A.Salem V.Carvalho E.Lawton Y.Li andJ. W.Stokes.2024. \u201cVLMGuard: Defending VLMs Against Malicious Prompts via Unlabeled Data.\u201darXiv preprint arXiv:2410.00296."},{"key":"e_1_2_8_8_1","doi-asserted-by":"crossref","unstructured":"Du X. G.Gozum Y.Ming andY.Li.2022. \u201cSIREN: Shaping Representations for Detecting Out\u2010of\u2010Distribution Objects.\u201d InAdvances in Neural Information Processing Systems.","DOI":"10.52202\/068431-1486"},{"key":"e_1_2_8_9_1","unstructured":"Du X. Y.Sun andY.Li.2024. \u201cWhen and How Does In\u2010Distribution Label Help Out\u2010of\u2010Distribution Detection?\u201d InInternational Conference on Machine Learning."},{"key":"e_1_2_8_10_1","doi-asserted-by":"crossref","unstructured":"Du X. Y.Sun X.Zhu andY.Li.2023. \u201cDream the Impossible: Outlier Imagination with Diffusion Models.\u201d InAdvances in Neural Information Processing Systems.","DOI":"10.52202\/075280-2660"},{"key":"e_1_2_8_11_1","doi-asserted-by":"crossref","unstructured":"Du X. X.Wang G.Gozum andY.Li.2022. \u201cUnknown\u2010Aware Object Detection: Learning What You Don't Know from Videos in the Wild.\u201d InProceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition.","DOI":"10.1109\/CVPR52688.2022.01331"},{"key":"e_1_2_8_12_1","unstructured":"Du X. Z.Wang M.Cai andY.Li.2022. \u201cVOS: Learning What You Don't Know by Virtual Outlier Synthesis.\u201d InInternational Conference on Learning Representations."},{"key":"e_1_2_8_13_1","doi-asserted-by":"crossref","unstructured":"Du X. C.Xiao andY.Li.2024. \u201cHaloScope: Harnessing Unlabeled LLM Generations for Hallucination Detection.\u201d InAdvances in Neural Information Processing Systems.","DOI":"10.52202\/079017-3270"},{"key":"e_1_2_8_14_1","doi-asserted-by":"crossref","unstructured":"Du X. D.Zhong andP.Li.2019. \u201cLow\u2010Shot Palmprint Recognition Based on Meta\u2010Siamese Network.\u201d In2019 IEEE International Conference on Multimedia and Expo (ICME) 79\u201384.IEEE.","DOI":"10.1109\/ICME.2019.00022"},{"key":"e_1_2_8_15_1","doi-asserted-by":"crossref","unstructured":"Du X. D.Zhong andH.Shao.2019a. \u201cBuilding an Active Palmprint Recognition System.\u201d In2019 IEEE International Conference on Image Processing (ICIP) 1685\u20131689.IEEE.","DOI":"10.1109\/ICIP.2019.8803135"},{"key":"e_1_2_8_16_1","doi-asserted-by":"crossref","unstructured":"Du X. D.Zhong andH.Shao.2019b. \u201cContinual Palmprint Recognition Without Forgetting.\u201d In2019 IEEE International Conference on Image Processing (ICIP) 1158\u20131162.IEEE.","DOI":"10.1109\/ICIP.2019.8803748"},{"key":"e_1_2_8_17_1","doi-asserted-by":"publisher","DOI":"10.1109\/TCSVT.2020.3024593"},{"key":"e_1_2_8_18_1","doi-asserted-by":"crossref","unstructured":"Gu J. Y.Ming Y.Zhou J.Kuen V.Morariu H.Zhao R.Zhang etal.2023. \u201cA Critical Analysis of Document Out\u2010of\u2010Distribution Detection.\u201d InFindings of the Association for Computational Linguistics: EMNLP 2023 4973\u20134999.","DOI":"10.18653\/v1\/2023.findings-emnlp.332"},{"key":"e_1_2_8_19_1","unstructured":"Hendrycks D. M.Mazeika andT.Dietterich.2019. \u201cDeep Anomaly Detection with Outlier Exposure.\u201d InInternational Conference on Learning Representations."},{"key":"e_1_2_8_20_1","doi-asserted-by":"crossref","unstructured":"Inkawhich M. N.Inkawhich H.Li andY.Chen.2024. \u201cTunable Hybrid Proposal Networks for the Open World.\u201d InProceedings of the IEEE\/CVF Winter Conference on Applications of Computer Vision 1988\u20131999.","DOI":"10.1109\/WACV57701.2024.00199"},{"key":"e_1_2_8_21_1","doi-asserted-by":"crossref","unstructured":"Lang H. Y.Zheng J.Sun F.Huang L.Si andY.Li.2022. \u201cEstimating Soft Labels for Out\u2010of\u2010Domain Intent Detection.\u201d InProceedings of the Conference on Empirical Methods in Natural Language Processing.","DOI":"10.18653\/v1\/2022.emnlp-main.18"},{"key":"e_1_2_8_22_1","unstructured":"Lee K. H.Lee K.Lee andJ.Shin.2018. \u201cTraining Confidence\u2010Calibrated Classifiers for Detecting Out\u2010of\u2010Distribution Samples.\u201d InInternational Conference on Learning Representations."},{"key":"e_1_2_8_23_1","doi-asserted-by":"crossref","unstructured":"Lin T.\u2010Y. M.Maire S.Belongie J.Hays P.Perona D.Ramanan P.Doll\u00e1r andC. L.Zitnick.2014. \u201cMicrosoft COCO: Common Objects in Context.\u201d InEuropean Conference on Computer Vision 740\u2013755.","DOI":"10.1007\/978-3-319-10602-1_48"},{"key":"e_1_2_8_24_1","doi-asserted-by":"crossref","unstructured":"Liu B. J. G.Boysen I. C.Unarta X.Du Y.Li andX.Huang.2025. \u201cExploring Transition States of Protein Conformational Changes via Out\u2010of\u2010Distribution Detection in the Hyperspherical Latent Space.\u201d Nature Communications.","DOI":"10.26434\/chemrxiv-2024-r8gjv"},{"key":"e_1_2_8_25_1","doi-asserted-by":"crossref","unstructured":"Liu J. X.Wen S.Zhao Y.Chen andX.Qi.2024. \u201cCan OOD Object Detectors Learn from Foundation Models?\u201d InEuropean Conference on Computer Vision.","DOI":"10.1007\/978-3-031-73254-6_13"},{"key":"e_1_2_8_26_1","unstructured":"Liu L. andY.Qin.2024. \u201cFast Decision Boundary Based Out\u2010of\u2010Distribution Detector.\u201d InInternational Conference on Machine Learning."},{"key":"e_1_2_8_27_1","unstructured":"Liu W. X.Wang J.Owens andY.Li.2020. \u201cEnergy\u2010Based Out\u2010of\u2010Distribution Detection.\u201d InAdvances in Neural Information Processing Systems."},{"key":"e_1_2_8_28_1","doi-asserted-by":"crossref","unstructured":"Narayanaswamy V. Y.Mubarka R.Anirudh D.Rajan andJ. J.Thiagarajan.2023. \u201cExploring Inlier and Outlier Specification for Improved Medical OOD Detection.\u201d InProceedings of the IEEE\/CVF International Conference on Computer Vision 4589\u20134598.","DOI":"10.1109\/ICCVW60793.2023.00493"},{"key":"e_1_2_8_29_1","doi-asserted-by":"crossref","unstructured":"Nguyen A. J.Yosinski andJ.Clune.2015. \u201cDeep Neural Networks Are Easily Fooled: High Confidence Predictions for Unrecognizable Images.\u201d InProceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition.","DOI":"10.1109\/CVPR.2015.7298640"},{"key":"e_1_2_8_30_1","doi-asserted-by":"crossref","unstructured":"Seifi S. D. O.Reino N.Chumerin andR.Aljundi.2024. \u201cOOD Aware Supervised Contrastive Learning.\u201d InProceedings of the IEEE\/CVF Winter Conference on Applications of Computer Vision 1956\u20131966.","DOI":"10.1109\/WACV57701.2024.00196"},{"key":"e_1_2_8_31_1","doi-asserted-by":"crossref","unstructured":"Shao H. D.Zhong andX.Du.2019. \u201cCross\u2010Domain Palmprint Recognition Based on Transfer Convolutional Autoencoder.\u201d In2019 IEEE International Conference on Image Processing (ICIP) 1153\u20131157.IEEE.","DOI":"10.1109\/ICIP.2019.8803778"},{"key":"e_1_2_8_32_1","unstructured":"Tao L. X.Du X.Zhu andY.Li.2023. \u201cNon\u2010Parametric Outlier Synthesis.\u201d InInternational Conference on Learning Representations."},{"key":"e_1_2_8_33_1","doi-asserted-by":"crossref","unstructured":"Um S. andJ. C.Ye.2024. \u201cSelf\u2010Guided Generation of Minority Samples Using Diffusion Models.\u201d InEuropean Conference on Computer Vision.","DOI":"10.1007\/978-3-031-73113-6_24"},{"key":"e_1_2_8_34_1","volume-title":"The Nature of Statistical Learning Theory","author":"Vapnik V","year":"1999"},{"key":"e_1_2_8_35_1","doi-asserted-by":"crossref","unstructured":"Yang J. P.Wang D.Zou Z.Zhou K.Ding W.Peng H.Wang et al.2022. \u201cOpenOOD: Benchmarking Generalized Out\u2010of\u2010Distribution Detection.\u201d InAdvances in Neural Information Processing Systems.","DOI":"10.52202\/068431-2362"},{"key":"e_1_2_8_36_1","unstructured":"Yeh M.\u2010H. J.Wang X.Du S.Park L.Tao S.Im andY.Li.2025. \u201cChallenges and Future Directions of Data\u2010Centric AI Alignment.\u201d InInternational Conference on Machine Learning."},{"key":"e_1_2_8_37_1","doi-asserted-by":"crossref","unstructured":"Yoon S. S.Yoon H.Lee Y. S.Sim S.Choi K.Lee H.\u2010S.Cho andW.Lim.2025. \u201cDiffusion based Semantic Outlier Generation via Nuisance Awareness for Out\u2010of\u2010Distribution Detection.\u201d InAAAI Conference on Artificial Intelligence.","DOI":"10.1609\/aaai.v39i12.33427"},{"key":"e_1_2_8_38_1","doi-asserted-by":"crossref","unstructured":"Yu F. H.Chen X.Wang W.Xian Y.Chen F.Liu V.Madhavan andT.Darrell.2020. \u201cBDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learning.\u201d InProceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition.","DOI":"10.1109\/CVPR42600.2020.00271"},{"key":"e_1_2_8_39_1","doi-asserted-by":"crossref","unstructured":"Zang Y. H.Goh J.Susskind andC.Huang.2024. \u201cOvercoming the Pitfalls of Vision\u2010Language Model Finetuning for OOD Generalization.\u201d InInternational Conference on Learning Representations.","DOI":"10.1145\/3664647.3680591"},{"key":"e_1_2_8_40_1","unstructured":"Zhang J. J.Yang P.Wang H.Wang Y.Lin H.Zhang Y.Sun et\u00a0al.2023. \u201cOpenOOD v1.5: Enhanced Benchmark for Out\u2010of\u2010Distribution Detection.\u201dJournal of Data\u2010centric Machine Learning Research."},{"key":"e_1_2_8_41_1","doi-asserted-by":"publisher","DOI":"10.1109\/TIFS.2019.2912552"}],"container-title":["AI Magazine"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/pdf\/10.1002\/aaai.70058","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/full-xml\/10.1002\/aaai.70058","content-type":"application\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/pdf\/10.1002\/aaai.70058","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,7]],"date-time":"2026-07-07T21:10:03Z","timestamp":1783458603000},"score":1,"resource":{"primary":{"URL":"https:\/\/onlinelibrary.wiley.com\/doi\/10.1002\/aaai.70058"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,4,9]]},"references-count":40,"journal-issue":{"issue":"2","published-print":{"date-parts":[[2026,6]]}},"alternative-id":["10.1002\/aaai.70058"],"URL":"https:\/\/doi.org\/10.1002\/aaai.70058","archive":["Portico"],"relation":{},"ISSN":["0738-4602","2371-9621"],"issn-type":[{"value":"0738-4602","type":"print"},{"value":"2371-9621","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,4,9]]},"assertion":[{"value":"2026-02-07","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-03-26","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-04-09","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}],"article-number":"e70058"}}