{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,8,18]],"date-time":"2026-08-18T01:47:04Z","timestamp":1787017624391,"version":"build-2736575974"},"reference-count":37,"publisher":"Association for Computing Machinery (ACM)","issue":"2","funder":[{"DOI":"10.13039\/501100001809","name":"National Natural Science Foundation of China","doi-asserted-by":"crossref","award":["62272046, 62132019, 62472033, 61872337"],"award-info":[{"award-number":["62272046, 62132019, 62472033, 61872337"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"crossref"}]},{"DOI":"10.13039\/501100006579","name":"Ministry of Industry and Information Technology","doi-asserted-by":"crossref","award":["CEIEC-20240"],"award-info":[{"award-number":["CEIEC-20240"]}],"id":[{"id":"10.13039\/501100006579","id-type":"DOI","asserted-by":"crossref"}]},{"name":"Science and Technology Innovation Project of Beijing Institute of Technology","award":["2023CX01017"],"award-info":[{"award-number":["2023CX01017"]}]},{"name":"Open Project of Key Laboratory of the Ministry of Education","award":["2023PY002"],"award-info":[{"award-number":["2023PY002"]}]}],"content-domain":{"domain":["dl.acm.org"],"crossmark-restriction":true},"short-container-title":["ACM Trans. Internet Things"],"published-print":{"date-parts":[[2026,5,31]]},"abstract":"<jats:p>Large vision and language models deployed at edge encounter continuously evolving input distributions, including not only new tasks but also highly unbalanced long-tail classes. For example, smart-surveillance cameras frequently capture common objects such as pedestrians and cars, while only occasionally observing tail classes like horseback riders or stroller pushers. However, most existing long-tail mitigation techniques are designed for fixed pretraining data. This often leads to poor accuracy of tail classes on evolving data and incurs high computational costs on edge devices. In this article, we propose EdgeTail, a lightweight long-tail mitigation method for edge-side continual learning. EdgeTail\u2019s key design features are: (i) optimal long-tailed mitigation solution search, which adaptively selects the best long-tail learning method for the current distribution\/task; and (ii) graph attention classifier and multi-branch adapter, which improves the quality and stability of tail class representations with small overheads. We implement EdgeTail in PyTorch and extensively evaluate it against state-of-the-art methods. The results show that EdgeTail improves the average accuracy by 36.09% under fixed training windows and by 31.12% under different training window sizes.<\/jats:p>","DOI":"10.1145\/3786774","type":"journal-article","created":{"date-parts":[[2025,12,25]],"date-time":"2025-12-25T12:07:34Z","timestamp":1766664454000},"page":"1-22","update-policy":"https:\/\/doi.org\/10.1145\/crossmark-policy","source":"Crossref","is-referenced-by-count":2,"title":["EdgeTail: Mitigating Long-Tail Visual Problems in Continual Learning at Edge"],"prefix":"10.1145","volume":"7","author":[{"ORCID":"https:\/\/orcid.org\/0009-0001-3889-8886","authenticated-orcid":false,"given":"Yuzhong","family":"Ouyang","sequence":"first","affiliation":[{"name":"Beijing Institute of Technology","place":["Haidian, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0001-8193-861X","authenticated-orcid":false,"given":"Xiaoning","family":"Wu","sequence":"additional","affiliation":[{"name":"Beijing Institute of Technology","place":["Beijing, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0001-5133-1462","authenticated-orcid":false,"given":"Menglin","family":"Yang","sequence":"additional","affiliation":[{"name":"Beijing Institute of Technology","place":["Beijing, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-6894-1921","authenticated-orcid":false,"given":"Rui","family":"Han","sequence":"additional","affiliation":[{"name":"Beijing Institute of Technology","place":["Beijing, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0009-0005-2791-0762","authenticated-orcid":false,"given":"Anjie","family":"Luo","sequence":"additional","affiliation":[{"name":"Beijing Institute of Technology","place":["Beijing, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-0252-329X","authenticated-orcid":false,"given":"Chi Harold","family":"Liu","sequence":"additional","affiliation":[{"name":"Beijing Institute of Technology","place":["Beijing, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5689-4742","authenticated-orcid":false,"given":"Jing","family":"Chen","sequence":"additional","affiliation":[{"name":"Key Laboratory of Computing Power Network and Information Security, Ministry of Education","place":["Jinan, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4262-874X","authenticated-orcid":false,"given":"Ying","family":"Guo","sequence":"additional","affiliation":[{"name":"Key Laboratory of Computing Power Network and Information Security, Ministry of Education","place":["Jinan, China"]}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"320","published-online":{"date-parts":[[2026,2,9]]},"reference":[{"key":"e_1_3_2_2_2","first-page":"95","volume-title":"Proceedings of the European Conference on Computer Vision","author":"Chou Hsin-Ping","year":"2020","unstructured":"Hsin-Ping Chou, Shih-Chieh Chang, Jia-Yu Pan, Wei Wei, and Da-Cheng Juan. 2020. Remix: Rebalanced mixup. In Proceedings of the European Conference on Computer Vision. Springer, 95\u2013110."},{"key":"e_1_3_2_3_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00949"},{"key":"e_1_3_2_4_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52729.2023.01518"},{"key":"e_1_3_2_5_2","first-page":"3762","volume-title":"Proceedings of the International Conference on Artificial Intelligence and Statistics","author":"Farajtabar Mehrdad","year":"2020","unstructured":"Mehrdad Farajtabar, Navid Azizan, Alex Mott, and Ang Li. 2020. Orthogonal gradient descent for continual learning. In Proceedings of the International Conference on Artificial Intelligence and Statistics. PMLR, 3762\u20133773."},{"key":"e_1_3_2_6_2","first-page":"60329","article-title":"Enhancing minority classes by mixing: An adaptative optimal transport approach for long-tailed classification","volume":"36","author":"Gao Jintong","year":"2023","unstructured":"Jintong Gao, He Zhao, Zhuo Li, and Dandan Guo. 2023. Enhancing minority classes by mixing: An adaptative optimal transport approach for long-tailed classification. Advances in Neural Information Processing Systems 36 (2023), 60329\u201360348.","journal-title":"Advances in Neural Information Processing Systems"},{"key":"e_1_3_2_7_2","first-page":"10","article-title":"Long-tailed continual learning for visual food recognition","author":"He Jiangpeng","year":"2025","unstructured":"Jiangpeng He, Xiaoyan Zhang, Luotao Lin, Jack Ma, Heather A. Eicher-Miller, and Fengqing Zhu. 2025. Long-tailed continual learning for visual food recognition. IEEE Transactions on Multimedia (2025), 10\u20131109.","journal-title":"IEEE Transactions on Multimedia"},{"key":"e_1_3_2_8_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.580"},{"key":"e_1_3_2_9_2","first-page":"2496","volume-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition Workshops","author":"Huang Zhiqiang","year":"2020","unstructured":"Zhiqiang Huang, Yunchao Zhang, Yutong Li, Zhihui Jiang, and Hang Gao. 2020. Counting: An efficient long-tailed recognition method. In Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition Workshops. 2496\u20132505."},{"key":"e_1_3_2_10_2","article-title":"Meta-learning representations for continual learning","volume":"32","author":"Javed Khurram","year":"2019","unstructured":"Khurram Javed and Martha White. 2019. Meta-learning representations for continual learning. Advances in Neural Information Processing Systems 32 (2019), 1818\u20131828.","journal-title":"Advances in Neural Information Processing Systems"},{"key":"e_1_3_2_11_2","unstructured":"Bingyi Kang Saining Xie Marcus Rohrbach Zhicheng Yan Albert Gordo Jiashi Feng and Yannis Kalantidis. 2019. Decoupling representation and classifier for long-tailed recognition. arXiv:1910.09217. Retrieved from https:\/\/arxiv.org\/abs\/1910.09217"},{"key":"e_1_3_2_12_2","unstructured":"Jared Kaplan Sam McCandlish Tom Henighan Tom B. Brown Benjamin Chess Rewon Child Scott Gray Alec Radford Jeffrey Wu and Dario Amodei. 2020. Scaling laws for neural language models. arXiv:2001.08361. Retrieved from https:\/\/arxiv.org\/abs\/2001.08361"},{"key":"e_1_3_2_13_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-58601-0_25"},{"key":"e_1_3_2_14_2","doi-asserted-by":"publisher","DOI":"10.1073\/pnas.1611835114"},{"key":"e_1_3_2_15_2","first-page":"9765","volume-title":"Proceedings of the 38th International Conference on Machine Learning (ICML)","author":"Li Chao","year":"2021","unstructured":"Chao Li, Xiujun Li, Yining Xia, Yonatan Bisk, Jianfeng Gao, Wen-tau Yih, Asli Celikyilmaz, Xuedong Song, Lijuan Gao, Kai-Wei Chang, and Luke Zettlemoyer. 2021. ALIGN: Scaling up visual and vision-language representation learning with noisy text supervision. In Proceedings of the 38th International Conference on Machine Learning (ICML). 9765\u20139775."},{"key":"e_1_3_2_16_2","volume-title":"Proceedings of the IEEE\/CVF International Conference on Computer Vision","author":"Liu Meng","year":"2021","unstructured":"Meng Liu, Hanwang Wu, Wenhan Wang, Yunchao Zhao, and Rongrong Ji. 2021. Video long-tailed recognition. In Proceedings of the IEEE\/CVF International Conference on Computer Vision. 2531\u20132540."},{"key":"e_1_3_2_17_2","first-page":"495","volume-title":"Proceedings of the European Conference on Computer Vision","author":"Liu Xialei","year":"2022","unstructured":"Xialei Liu, Yu-Song Hu, Xu-Sheng Cao, Andrew D. Bagdanov, Ke Li, and Ming-Ming Cheng. 2022. Long-tailed class incremental learning. In Proceedings of the European Conference on Computer Vision. Springer, 495\u2013512."},{"key":"e_1_3_2_18_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2019.00264"},{"key":"e_1_3_2_19_2","article-title":"Gradient episodic memory for continual learning","volume":"30","author":"Lopez-Paz David","year":"2017","unstructured":"David Lopez-Paz and Marc\u2019Aurelio Ranzato. 2017. Gradient episodic memory for continual learning. Advances in Neural Information Processing Systems 30 (2017), 6467\u20136476.","journal-title":"Advances in Neural Information Processing Systems"},{"key":"e_1_3_2_20_2","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-01216-8_12"},{"key":"e_1_3_2_21_2","first-page":"67","volume-title":"Proceedings of the European Conference on Computer Vision","author":"Mallya Arun","year":"2018","unstructured":"Arun Mallya, Dillon Davis, and Svetlana Lazebnik. 2018. Piggyback: Adapting a single network to multiple tasks by learning to mask weights. In Proceedings of the European Conference on Computer Vision. 67\u201382."},{"key":"e_1_3_2_22_2","unstructured":"Aditya Krishna Menon Sadeep Jayasumana Ankit Singh Rawat Himanshu Jain Andreas Veit and Sanjiv Kumar. 2020. Long-tail learning via logit adjustment. arXiv:2007.07314. Retrieved from https:\/\/arxiv.org\/abs\/2007.07314"},{"key":"e_1_3_2_23_2","doi-asserted-by":"publisher","DOI":"10.19678\/j.issn.1000-3428.0069287"},{"key":"e_1_3_2_24_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52688.2022.00676"},{"key":"e_1_3_2_25_2","first-page":"4054","volume-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition","author":"Raghavan Siddeshwar","year":"2024","unstructured":"Siddeshwar Raghavan, Jiangpeng He, and Fengqing Zhu. 2024. DELTA: Decoupling long-tailed online continual learning. In Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition. 4054\u20134064."},{"key":"e_1_3_2_26_2","article-title":"Prototypical networks for few-shot learning","volume":"30","author":"Snell Jake","year":"2017","unstructured":"Jake Snell, Kevin Swersky, and Richard Zemel. 2017. Prototypical networks for few-shot learning. Advances in Neural Information Processing Systems 30 (2017), 4080\u20134090.","journal-title":"Advances in Neural Information Processing Systems"},{"key":"e_1_3_2_27_2","first-page":"4223","volume-title":"Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition","author":"Teney Damien","year":"2018","unstructured":"Damien Teney, Peter Anderson, Xiaodong He, and Anton van den Hengel. 2018. Tips and tricks for visual question answering: Learnings from the 2017 challenge. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 4223\u20134232."},{"key":"e_1_3_2_28_2","unstructured":"Petar Veli\u010dkovi\u0107 Guillem Cucurull Arantxa Casanova Adriana Romero Pietro Lio and Yoshua Bengio. 2017. Graph attention networks. arXiv:1710.10903. Retrieved from https:\/\/arxiv.org\/abs\/1710.10903"},{"key":"e_1_3_2_29_2","unstructured":"Eli Verwimp Rahaf Aljundi Shai Ben-David Matthias Bethge Andrea Cossu Alexander Gepperth Tyler L Hayes Eyke H\u00fcllermeier Christopher Kanan Dhireesha Kudithipudi et\u00a0al. 2023. Continual learning: Applications and the road forward. arXiv:2311.11908. Retrieved from https:\/\/arxiv.org\/abs\/2311.11908"},{"key":"e_1_3_2_30_2","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2024.3498346"},{"key":"e_1_3_2_31_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.02702"},{"key":"e_1_3_2_32_2","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2019.00512"},{"key":"e_1_3_2_33_2","article-title":"Learning to model the tail","volume":"30","author":"Wang Yu-Xiong","year":"2017","unstructured":"Yu-Xiong Wang, Deva Ramanan, and Martial Hebert. 2017. Learning to model the tail. Advances in Neural Information Processing Systems 30 (2017), 7029\u20137039.","journal-title":"Advances in Neural Information Processing Systems"},{"key":"e_1_3_2_34_2","volume-title":"Proceedings of the European Conference on Computer Vision","author":"Wu Ji","year":"2022","unstructured":"Ji Wu, Yaliang Zheng, Zhiqiang Zhou, and Yujiu Yang. 2022. Contrastive prototype learning for long-tailed recognition. In Proceedings of the European Conference on Computer Vision. 341\u2013357."},{"key":"e_1_3_2_35_2","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-022-01622-8"},{"key":"e_1_3_2_36_2","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR52733.2024.02191"},{"key":"e_1_3_2_37_2","first-page":"6489","volume-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition","author":"Zhang Pengchuan","year":"2021","unstructured":"Pengchuan Zhang, Xiujun Li, Xiaowei Hu, Jianwei Yang, Lei Zhang, Lijuan Wang, Yejin Choi, and Jianfeng Gao. 2021. VinVL: Making visual representations matter in vision-language models. In Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition. 6489\u20136499."},{"key":"e_1_3_2_38_2","doi-asserted-by":"publisher","DOI":"10.1109\/TPAMI.2023.3268118"}],"container-title":["ACM Transactions on Internet of Things"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dl.acm.org\/doi\/pdf\/10.1145\/3786774","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,2,9]],"date-time":"2026-02-09T13:23:51Z","timestamp":1770643431000},"score":1,"resource":{"primary":{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3786774"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,2,9]]},"references-count":37,"journal-issue":{"issue":"2","published-print":{"date-parts":[[2026,5,31]]}},"alternative-id":["10.1145\/3786774"],"URL":"https:\/\/doi.org\/10.1145\/3786774","relation":{},"ISSN":["2691-1914","2577-6207"],"issn-type":[{"value":"2691-1914","type":"print"},{"value":"2577-6207","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,2,9]]},"assertion":[{"value":"2025-09-13","order":0,"name":"received","label":"Received","group":{"name":"publication_history","label":"Publication History"}},{"value":"2025-12-15","order":2,"name":"accepted","label":"Accepted","group":{"name":"publication_history","label":"Publication History"}},{"value":"2026-02-09","order":3,"name":"published","label":"Published","group":{"name":"publication_history","label":"Publication History"}}]}}