{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,5,19]],"date-time":"2026-05-19T07:12:20Z","timestamp":1779174740619,"version":"3.51.4"},"reference-count":100,"publisher":"Association for Computing Machinery (ACM)","issue":"3","funder":[{"DOI":"10.13039\/100000180","name":"Department of Homeland Security","doi-asserted-by":"crossref","award":["17STQAC00001-07-00"],"award-info":[{"award-number":["17STQAC00001-07-00"]}],"id":[{"id":"10.13039\/100000180","id-type":"DOI","asserted-by":"crossref"}]},{"DOI":"10.13039\/501100006374","name":"National Science Foundation","doi-asserted-by":"publisher","award":["2144923 and 2431532"],"award-info":[{"award-number":["2144923 and 2431532"]}],"id":[{"id":"10.13039\/501100006374","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["Proc. ACM Manag. Data"],"published-print":{"date-parts":[[2025,6,17]]},"abstract":"<jats:p>With the growing adoption of privacy-preserving machine learning algorithms, such as Differentially Private Stochastic Gradient Descent (DP-SGD), training or fine-tuning models on private datasets has become increasingly prevalent. This shift has led to the need for models offering varying privacy guarantees and utility levels to satisfy diverse user requirements. Managing numerous versions of large models introduces significant operational challenges, including increased inference latency, higher resource consumption, and elevated costs. Model deduplication is a technique widely used by many model serving and database systems to support high-performance and low-cost inference queries and model diagnosis queries. However, none of the existing model deduplication works has considered privacy, leading to unbounded aggregation of privacy costs for certain deduplicated models and inefficiencies when applied to deduplicate DP-trained models.<\/jats:p>\n                  <jats:p>We formalize the problem of deduplicating DP-trained models for the first time and propose a novel privacy- and accuracy-aware deduplication mechanism to address the problem. We developed a greedy strategy to select and assign base models to target models to minimize storage and privacy costs. When deduplicating a target model, we dynamically schedule accuracy validations and apply the Sparse Vector Technique to reduce the privacy costs associated with private validation data. Compared to baselines, our approach improved the compression ratio by up to 35\u00d7 for individual models (including large language models and vision transformers). We also observed up to 43\u00d7 inference speedup due to the reduction of I\/O operations.<\/jats:p>","DOI":"10.1145\/3725340","type":"journal-article","created":{"date-parts":[[2025,6,18]],"date-time":"2025-06-18T21:23:29Z","timestamp":1750281809000},"page":"1-28","source":"Crossref","is-referenced-by-count":2,"title":["Privacy and Accuracy-Aware AI\/ML Model Deduplication"],"prefix":"10.1145","volume":"3","author":[{"ORCID":"https:\/\/orcid.org\/0000-0003-2781-4242","authenticated-orcid":false,"given":"Hong","family":"Guan","sequence":"first","affiliation":[{"name":"Arizona State University, Tempe, USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5968-0344","authenticated-orcid":false,"given":"Lei","family":"Yu","sequence":"additional","affiliation":[{"name":"Rensselaer Polytechnic Institute, Troy, USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8481-3292","authenticated-orcid":false,"given":"Lixi","family":"Zhou","sequence":"additional","affiliation":[{"name":"Arizona State University, Tempe, USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7354-0428","authenticated-orcid":false,"given":"Li","family":"Xiong","sequence":"additional","affiliation":[{"name":"Emory University, Atlanta, USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0009-0004-6528-9324","authenticated-orcid":false,"given":"Kanchan","family":"Chowdhury","sequence":"additional","affiliation":[{"name":"Arizona State University, Tempe, USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-8902-222X","authenticated-orcid":false,"given":"Lulu","family":"Xie","sequence":"additional","affiliation":[{"name":"Arizona State University, Tempe, USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-4797-4294","authenticated-orcid":false,"given":"Xusheng","family":"Xiao","sequence":"additional","affiliation":[{"name":"Arizona State University, Tempe, USA"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-9849-0711","authenticated-orcid":false,"given":"Jia","family":"Zou","sequence":"additional","affiliation":[{"name":"Arizona State University, Tempe, USA"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"320","published-online":{"date-parts":[[2025,6,18]]},"reference":[{"key":"e_1_2_2_1_1","first-page":"308","article-title":"Deep learning with differential privacy","author":"Abadi Martin","year":"2016","unstructured":"Martin Abadi, Andy Chu, Ian Goodfellow, H Brendan McMahan, Ilya Mironov, Kunal Talwar, and Li Zhang. 2016. Deep learning with differential privacy. In CCS. 308-318.","journal-title":"CCS."},{"key":"e_1_2_2_2_1","doi-asserted-by":"publisher","DOI":"10.1016\/B978-155860869-6\/50058-5"},{"key":"e_1_2_2_3_1","doi-asserted-by":"publisher","DOI":"10.1136\/jamia.2009.000026"},{"key":"e_1_2_2_4_1","volume-title":"Proceedings of the Conference on Very Large Databases.","author":"Bernau Daniel","year":"2021","unstructured":"Daniel Bernau, G\u00fcnther Eibl, Philip W Grassal, Hannah Keller, and Florian Kerschbaum. 2021. Quantifying Identifiability to Choose and Audit o in Differentially Private Deep Learning. In Proceedings of the Conference on Very Large Databases."},{"key":"e_1_2_2_5_1","volume-title":"Jacob Portes, Mansheej Paul, Philip Greengard, Connor Jennings, Daniel King, Sam Havens, Vitaliy Chiley, Jonathan Frankle, et al.","author":"Biderman Dan","year":"2024","unstructured":"Dan Biderman, Jose Gonzalez Ortiz, Jacob Portes, Mansheej Paul, Philip Greengard, Connor Jennings, Daniel King, Sam Havens, Vitaliy Chiley, Jonathan Frankle, et al., 2024. Lora learns less and forgets less. arXiv preprint arXiv:2405.09673, (2024)."},{"key":"e_1_2_2_6_1","doi-asserted-by":"publisher","DOI":"10.1109\/ICDM.2006.13"},{"key":"e_1_2_2_7_1","first-page":"129","article-title":"What is the state of neural network pruning","volume":"2","author":"Blalock Davis","year":"2020","unstructured":"Davis Blalock, Jose Javier Gonzalez Ortiz, Jonathan Frankle, and John Guttag. 2020. What is the state of neural network pruning? Proceedings of machine learning and systems, Vol. 2 (2020), 129-146.","journal-title":"Proceedings of machine learning and systems"},{"key":"e_1_2_2_8_1","doi-asserted-by":"publisher","DOI":"10.14778\/3007263.3007279"},{"key":"e_1_2_2_9_1","volume-title":"Scalable Blocking for Very Large Databases. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases. Springer, 303-319","author":"Borthwick Andrew","year":"2020","unstructured":"Andrew Borthwick, Stephen Ash, Bin Pang, Shehzad Qureshi, and Timothy Jones. 2020. Scalable Blocking for Very Large Databases. In Joint European Conference on Machine Learning and Knowledge Discovery in Databases. Springer, 303-319."},{"key":"e_1_2_2_10_1","first-page":"21","volume-title":"Proceedings. Compression and Complexity of SEQUENCES 1997 (Cat. No. 97TB100171)","author":"Broder Andrei Z","year":"1997","unstructured":"Andrei Z Broder. 1997. On the resemblance and containment of documents. In Proceedings. Compression and Complexity of SEQUENCES 1997 (Cat. No. 97TB100171). IEEE, 21-29."},{"key":"e_1_2_2_11_1","volume-title":"Differentially Private Bias-Term Fine-tuning of Foundation Models. In Forty-first International Conference on Machine Learning.","author":"Bu Zhiqi","unstructured":"Zhiqi Bu, Yu-Xiang Wang, Sheng Zha, and George Karypis. [n.d.]. Differentially Private Bias-Term Fine-tuning of Foundation Models. In Forty-first International Conference on Machine Learning."},{"key":"e_1_2_2_12_1","volume-title":"Differentially Private Bias-Term Fine-tuning of Foundation Models. In Workshop on Trustworthy and Socially Responsible Machine Learning, NeurIPS","author":"Bu Zhiqi","year":"2022","unstructured":"Zhiqi Bu, Yu-Xiang Wang, Sheng Zha, and George Karypis. 2022. Differentially Private Bias-Term Fine-tuning of Foundation Models. In Workshop on Trustworthy and Socially Responsible Machine Learning, NeurIPS 2022."},{"key":"e_1_2_2_13_1","doi-asserted-by":"publisher","DOI":"10.1016\/0377-2217(94)90338-7"},{"key":"e_1_2_2_14_1","unstructured":"Guillaume Maurice Jean-Bernard Chaslot Chaslot. 2010. Monte-carlo tree search. (2010)."},{"key":"e_1_2_2_15_1","doi-asserted-by":"publisher","DOI":"10.1145\/3299869.3300078"},{"key":"e_1_2_2_16_1","unstructured":"Lingjiao Chen Arun Kumar Jeffrey Naughton and Jignesh M Patel. 2016. Towards linear algebra over normalized data. arXiv preprint arXiv:1612.07448 (2016)."},{"key":"e_1_2_2_17_1","doi-asserted-by":"publisher","DOI":"10.14778\/2983200.2983203"},{"key":"e_1_2_2_18_1","volume-title":"Restrict data access using analysis rules. https:\/\/cloud.google.com\/bigquery\/docs\/analysis-rules Retrieved","author":"Cloud Google","year":"2024","unstructured":"Google Cloud. [n.d.]. Restrict data access using analysis rules. https:\/\/cloud.google.com\/bigquery\/docs\/analysis-rules Retrieved Jan 17, 2024 from"},{"key":"e_1_2_2_19_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.ipl.2006.06.003"},{"key":"e_1_2_2_20_1","unstructured":"Daniel Crankshaw. 2019. The Design and Implementation of Low-Latency Prediction Serving Systems. Ph.D. Dissertation. UC Berkeley."},{"key":"e_1_2_2_21_1","volume-title":"Clipper: A low-latency online prediction serving system. In 14th {USENIX} Symposium on Networked Systems Design and Implementation ({NSDI} 17). 613-627.","author":"Crankshaw Daniel","year":"2017","unstructured":"Daniel Crankshaw, Xin Wang, Guilio Zhou, Michael J Franklin, Joseph E Gonzalez, and Ion Stoica. 2017. Clipper: A low-latency online prediction serving system. In 14th {USENIX} Symposium on Networked Systems Design and Implementation ({NSDI} 17). 613-627."},{"key":"e_1_2_2_22_1","doi-asserted-by":"publisher","DOI":"10.1145\/997817.997857"},{"key":"e_1_2_2_23_1","unstructured":"Damien Desfontaines. 2021. A list of real-world uses of differential privacy. Ted is writing things (2021)."},{"key":"e_1_2_2_24_1","doi-asserted-by":"publisher","DOI":"10.1145\/2001576.2001710"},{"key":"e_1_2_2_25_1","volume-title":"International Conference on Learning Representations. https:\/\/openreview.net\/forum?id=YicbFdNTTy","author":"Dosovitskiy Alexey","year":"2021","unstructured":"Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkoreit, and Neil Houlsby. 2021. An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale. In International Conference on Learning Representations. https:\/\/openreview.net\/forum?id=YicbFdNTTy"},{"key":"e_1_2_2_26_1","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-540-79228-4_1"},{"key":"e_1_2_2_27_1","doi-asserted-by":"publisher","DOI":"10.1561\/0400000042"},{"key":"e_1_2_2_28_1","doi-asserted-by":"publisher","DOI":"10.1109\/TKDE.2007.250581"},{"key":"e_1_2_2_29_1","first-page":"4475","article-title":"Optimal brain compression: A framework for accurate post-training quantization and pruning","volume":"35","author":"Frantar Elias","year":"2022","unstructured":"Elias Frantar and Dan Alistarh. 2022. Optimal brain compression: A framework for accurate post-training quantization and pruning. Advances in Neural Information Processing Systems, Vol. 35 (2022), 4475-4488.","journal-title":"Advances in Neural Information Processing Systems"},{"key":"e_1_2_2_30_1","volume-title":"Gptq: Accurate post-training quantization for generative pre-trained transformers. arXiv preprint arXiv:2210.17323","author":"Frantar Elias","year":"2022","unstructured":"Elias Frantar, Saleh Ashkboos, Torsten Hoefler, and Dan Alistarh. 2022. Gptq: Accurate post-training quantization for generative pre-trained transformers. arXiv preprint arXiv:2210.17323, (2022)."},{"key":"e_1_2_2_31_1","doi-asserted-by":"publisher","DOI":"10.1145\/2810103.2813677"},{"key":"e_1_2_2_32_1","volume-title":"Dpsur: Accelerating differentially private stochastic gradient descent using selective update and release. arXiv preprint arXiv:2311.14056","author":"Fu Jie","year":"2023","unstructured":"Jie Fu, Qingqing Ye, Haibo Hu, Zhili Chen, Lulu Wang, Kuncan Wang, and Ran Xun. 2023. Dpsur: Accelerating differentially private stochastic gradient descent using selective update and release. arXiv preprint arXiv:2311.14056, (2023)."},{"key":"e_1_2_2_33_1","volume-title":"Deep learning with label differential privacy. Advances in neural information processing systems","author":"Ghazi Badih","year":"2021","unstructured":"Badih Ghazi, Noah Golowich, Ravi Kumar, Pasin Manurangsi, and Chiyuan Zhang. 2021. Deep learning with label differential privacy. Advances in neural information processing systems, Vol. 34 (2021), 27131-27145."},{"key":"e_1_2_2_34_1","doi-asserted-by":"publisher","DOI":"10.1201\/9781003162810-13"},{"key":"e_1_2_2_35_1","doi-asserted-by":"publisher","DOI":"10.1007\/s11263-021-01453-z"},{"key":"e_1_2_2_36_1","doi-asserted-by":"publisher","DOI":"10.1145\/3620678.3624656"},{"key":"e_1_2_2_37_1","unstructured":"Andrew Hard Kanishka Rao Rajiv Mathews Swaroop Ramaswamy Fran\u00e7oise Beaufays Sean Augenstein Hubert Eichner Chlo\u00e9 Kiddon and Daniel Ramage. 2018. Federated learning for mobile keyboard prediction. arXiv preprint arXiv:1811.03604 (2018)."},{"key":"e_1_2_2_38_1","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR.2016.90"},{"key":"e_1_2_2_39_1","volume-title":"Eugene Fratkin, Aleksander Gorajek, Kee Siong Ng, Caleb Welton, Xixuan Feng, Kun Li, et al.","author":"Hellerstein Joe","year":"2012","unstructured":"Joe Hellerstein, Christopher R\u00e9, Florian Schoppmann, Daisy Zhe Wang, Eugene Fratkin, Aleksander Gorajek, Kee Siong Ng, Caleb Welton, Xixuan Feng, Kun Li, et al., 2012. The MADlib analytics library or MAD skills, the SQL. arXiv preprint arXiv:1208.4165, (2012)."},{"key":"e_1_2_2_40_1","doi-asserted-by":"publisher","DOI":"10.1145\/568271.223807"},{"key":"e_1_2_2_41_1","volume-title":"Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685","author":"Hu Edward J","year":"2021","unstructured":"Edward J Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2021. Lora: Low-rank adaptation of large language models. arXiv preprint arXiv:2106.09685, (2021)."},{"key":"e_1_2_2_42_1","doi-asserted-by":"publisher","DOI":"10.1145\/3523273"},{"key":"e_1_2_2_43_1","doi-asserted-by":"publisher","DOI":"10.1145\/3581791.3596852"},{"key":"e_1_2_2_44_1","volume-title":"Chiron: Privacy-preserving machine learning as a service. arXiv preprint arXiv:1803.05961","author":"Hunt Tyler","year":"2018","unstructured":"Tyler Hunt, Congzheng Song, Reza Shokri, Vitaly Shmatikov, and Emmett Witchel. 2018. Chiron: Privacy-preserving machine learning as a service. arXiv preprint arXiv:1803.05961, (2018)."},{"key":"e_1_2_2_45_1","first-page":"22205","article-title":"Auditing differentially private machine learning: How private is private SGD","volume":"33","author":"Jagielski Matthew","year":"2020","unstructured":"Matthew Jagielski, Jonathan Ullman, and Alina Oprea. 2020. Auditing differentially private machine learning: How private is private SGD? Advances in Neural Information Processing Systems, Vol. 33 (2020), 22205-22216.","journal-title":"Advances in Neural Information Processing Systems"},{"key":"e_1_2_2_46_1","volume-title":"Proceedings of the VLDB Endowment","volume":"12","author":"Jankov Dimitrije","unstructured":"Dimitrije Jankov, Shangyu Luo, Binhang Yuan, Zhuhua Cai, Jia Zou, Chris Jermaine, and Zekai J Gao. [n.d.]. Declarative Recursive Computation on an RDBMS. Proceedings of the VLDB Endowment, Vol. 12, 7 ( [n.,d.])."},{"key":"e_1_2_2_47_1","doi-asserted-by":"publisher","DOI":"10.14778\/2367502.2367527"},{"key":"e_1_2_2_48_1","doi-asserted-by":"publisher","DOI":"10.1109\/ICDE.2012.22"},{"key":"e_1_2_2_49_1","unstructured":"Alex Krizhevsky. 2009. Learning multiple layers of features from tiny images. Technical Report."},{"key":"e_1_2_2_50_1","unstructured":"Alex Krizhevsky Geoffrey Hinton et al. 2009. Learning multiple layers of features from tiny images. (2009)."},{"key":"e_1_2_2_51_1","doi-asserted-by":"publisher","DOI":"10.1145\/2723372.2723713"},{"key":"e_1_2_2_52_1","volume-title":"International Conference on Learning Representations. https:\/\/openreview.net\/forum?id=HJeTo2VFwH","author":"Lee Namhoon","unstructured":"Namhoon Lee, Thalaiyasingam Ajanthan, Stephen Gould, and Philip H. S. Torr. 2020. A Signal Propagation Perspective for Pruning Neural Networks at Initialization. In International Conference on Learning Representations. https:\/\/openreview.net\/forum?id=HJeTo2VFwH"},{"key":"e_1_2_2_53_1","doi-asserted-by":"publisher","DOI":"10.1145\/3386901.3388947"},{"key":"e_1_2_2_54_1","volume-title":"Pruning Filters for Efficient ConvNets. In International Conference on Learning Representations. https:\/\/openreview.net\/forum?id=rJqFGTslg","author":"Li Hao","year":"2017","unstructured":"Hao Li, Asim Kadav, Igor Durdanovic, Hanan Samet, and Hans Peter Graf. 2017. Pruning Filters for Efficient ConvNets. In International Conference on Learning Representations. https:\/\/openreview.net\/forum?id=rJqFGTslg"},{"key":"e_1_2_2_55_1","doi-asserted-by":"publisher","DOI":"10.14778\/2732939.2732948"},{"key":"e_1_2_2_56_1","volume-title":"AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration. In MLSys.","author":"Lin Ji","year":"2024","unstructured":"Ji Lin, Jiaming Tang, Haotian Tang, Shang Yang, Wei-Ming Chen, Wei-Chen Wang, Guangxuan Xiao, Xingyu Dang, Chuang Gan, and Song Han. 2024. AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration. In MLSys."},{"key":"e_1_2_2_57_1","doi-asserted-by":"publisher","DOI":"10.14778\/3632093.3632095"},{"key":"e_1_2_2_58_1","doi-asserted-by":"publisher","DOI":"10.14778\/3447689.3447700"},{"key":"e_1_2_2_59_1","volume-title":"International Conference on Machine Learning. PMLR, 7021-7032","author":"Liu Liyang","year":"2021","unstructured":"Liyang Liu, Shilong Zhang, Zhanghui Kuang, Aojun Zhou, Jing-Hao Xue, Xinjiang Wang, Yimin Chen, Wenming Yang, Qingmin Liao, and Wayne Zhang. 2021b. Group fisher pruning for practical network compression. In International Conference on Machine Learning. PMLR, 7021-7032."},{"key":"e_1_2_2_60_1","volume-title":"Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692","author":"Liu Yinhan","year":"2019","unstructured":"Yinhan Liu, Myle Ott, Naman Goyal, Jingfei Du, Mandar Joshi, Danqi Chen, Omer Levy, Mike Lewis, Luke Zettlemoyer, and Veselin Stoyanov. 2019. Roberta: A robustly optimized bert pretraining approach. arXiv preprint arXiv:1907.11692, (2019)."},{"key":"e_1_2_2_61_1","doi-asserted-by":"publisher","DOI":"10.1109\/ICCV.2015.425"},{"key":"e_1_2_2_62_1","volume-title":"15th {USENIX} Symposium on Operating Systems Design and Implementation ({OSDI} 21). 55-74.","author":"Luo Tao","unstructured":"Tao Luo, Mingen Pan, Pierre Tholoniat, Asaf Cidon, Roxana Geambasu, and Mathias L\u00e9cuyer. 2021. Privacy budget scheduling. In 15th {USENIX} Symposium on Operating Systems Design and Implementation ({OSDI} 21). 55-74."},{"key":"e_1_2_2_63_1","doi-asserted-by":"publisher","DOI":"10.14778\/3055330.3055331"},{"key":"e_1_2_2_64_1","doi-asserted-by":"publisher","DOI":"10.5555\/2002472.2002491"},{"key":"e_1_2_2_65_1","doi-asserted-by":"publisher","DOI":"10.1609\/aaai.v31i1.10989"},{"key":"e_1_2_2_66_1","doi-asserted-by":"publisher","DOI":"10.1145\/1557019.1557090"},{"key":"e_1_2_2_67_1","doi-asserted-by":"publisher","DOI":"10.1145\/1559845.1559850"},{"key":"e_1_2_2_68_1","doi-asserted-by":"publisher","DOI":"10.1145\/1559845.1559850"},{"key":"e_1_2_2_69_1","volume-title":"Proceedings of the VLDB Endowment","volume":"11","author":"Tang Joty","year":"2018","unstructured":"Joty MESTS and MON Tang. 2018. Distributed representations of tuples for entity resolution. Proceedings of the VLDB Endowment, Vol. 11, 11 (2018)."},{"key":"e_1_2_2_70_1","first-page":"1613","volume-title":"32nd USENIX Security Symposium (USENIX Security 23)","author":"Nanayakkara Priyanka","year":"2023","unstructured":"Priyanka Nanayakkara, Mary Anne Smart, Rachel Cummings, Gabriel Kaptchuk, and Elissa M Redmiles. 2023. What are the chances? explaining the epsilon parameter in differential privacy. In 32nd USENIX Security Symposium (USENIX Security 23). 1613-1630."},{"key":"e_1_2_2_71_1","unstructured":"Yuval Netzer Tao Wang Adam Coates Alessandro Bissacco Bo Wu and Andrew Y Ng. 2011. Reading digits in natural images with unsupervised feature learning. (2011)."},{"key":"e_1_2_2_72_1","unstructured":"S Ramaswamy R Mathews K Rao and F Beaufays. 2019. Learning for Emoji Prediction in a Mobile Keyboard. arXiv preprint arXiv:1906.04329 (2019)."},{"key":"e_1_2_2_73_1","volume-title":"International Conference on Machine Learning. PMLR, 29354-29387","author":"Rust Phillip","year":"2023","unstructured":"Phillip Rust and Anders S\u00f8gaard. 2023. Differential privacy, linguistic fairness, and training data influence: Impossibility and possibility theorems for multilingual language models. In International Conference on Machine Learning. PMLR, 29354-29387."},{"key":"e_1_2_2_74_1","doi-asserted-by":"publisher","DOI":"10.1145\/3341301.3359658"},{"key":"e_1_2_2_75_1","doi-asserted-by":"publisher","DOI":"10.1109\/SP.2017.41"},{"key":"e_1_2_2_76_1","volume-title":"Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps. CoRR","author":"Simonyan Karen","year":"2013","unstructured":"Karen Simonyan, Andrea Vedaldi, and Andrew Zisserman. 2013. Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps. CoRR, Vol. abs\/1312.6034 (2013). https:\/\/api.semanticscholar.org\/CorpusID:1450294"},{"key":"e_1_2_2_77_1","unstructured":"snowflake. [n.d.]. Differential Privacy in Snowflake. https:\/\/docs.snowflake.com\/en\/user-guide\/diff-privacy\/differential-privacy-overview#label-diff-privacy-loss-budgets."},{"key":"e_1_2_2_78_1","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D13-1170"},{"key":"e_1_2_2_79_1","doi-asserted-by":"publisher","DOI":"10.1016\/j.neunet.2012.02.016"},{"key":"e_1_2_2_80_1","volume-title":"The Twelfth International Conference on Learning Representations. https:\/\/openreview.net\/forum?id=PxoFut3dWW","author":"Sun Mingjie","year":"2024","unstructured":"Mingjie Sun, Zhuang Liu, Anna Bair, and J Zico Kolter. 2024. A Simple and Effective Pruning Approach for Large Language Models. In The Twelfth International Conference on Learning Representations. https:\/\/openreview.net\/forum?id=PxoFut3dWW"},{"key":"e_1_2_2_81_1","doi-asserted-by":"publisher","DOI":"10.14778\/3476249.3476294"},{"key":"e_1_2_2_82_1","volume-title":"International Conference on Learning Representations. https:\/\/openreview.net\/forum?id=YTWGvpFOQD-","author":"Tramer Florian","year":"2021","unstructured":"Florian Tramer and Dan Boneh. 2021. Differentially Private Learning Needs Better Features (or Much More Data). In International Conference on Learning Representations. https:\/\/openreview.net\/forum?id=YTWGvpFOQD-"},{"key":"e_1_2_2_83_1","doi-asserted-by":"publisher","DOI":"10.1145\/3183713.3196934"},{"key":"e_1_2_2_84_1","volume-title":"GLUE: A multi-task benchmark and analysis platform for natural language understanding. arXiv preprint arXiv:1804.07461","author":"Wang Alex","year":"2018","unstructured":"Alex Wang, Amanpreet Singh, Julian Michael, Felix Hill, Omer Levy, and Samuel R Bowman. 2018. GLUE: A multi-task benchmark and analysis platform for natural language understanding. arXiv preprint arXiv:1804.07461, (2018)."},{"key":"e_1_2_2_85_1","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/N18--1101"},{"key":"e_1_2_2_86_1","doi-asserted-by":"publisher","DOI":"10.14778\/1453856.1453957"},{"key":"e_1_2_2_87_1","unstructured":"Zheng Xu and Yanxiang Zhang. 2017. Advances in private training for production on-device language models."},{"key":"e_1_2_2_88_1","unstructured":"Enneng Yang Li Shen Guibing Guo Xingwei Wang Xiaochun Cao Jie Zhang and Dacheng Tao. 2024. Model merging in llms mllms and beyond: Methods theories applications and opportunities. arXiv preprint arXiv:2408.07666 (2024)."},{"key":"e_1_2_2_89_1","unstructured":"Wangsong Yin Mengwei Xu Yuanchun Li and Xuanzhe Liu. 2024. Llm as a system service on mobile devices. arXiv preprint arXiv:2403.11805 (2024)."},{"key":"e_1_2_2_90_1","doi-asserted-by":"publisher","DOI":"10.1109\/TKDE.2016.2638838"},{"key":"e_1_2_2_91_1","unstructured":"Da Yu Huishuai Zhang Wei Chen and Tie-Yan Liu. 2021a. Do not let privacy overbill utility: Gradient embedding perturbation for private learning. arXiv preprint arXiv:2102.12677 (2021)."},{"key":"e_1_2_2_92_1","volume-title":"International Conference on Machine Learning. PMLR, 12208-12218","author":"Yu Da","year":"2021","unstructured":"Da Yu, Huishuai Zhang, Wei Chen, Jian Yin, and Tie-Yan Liu. 2021b. Large scale private learning via low-rank reparametrization. In International Conference on Machine Learning. PMLR, 12208-12218."},{"key":"e_1_2_2_93_1","doi-asserted-by":"publisher","DOI":"10.14778\/3457390.3457399"},{"key":"e_1_2_2_94_1","doi-asserted-by":"publisher","DOI":"10.1145\/3626761"},{"key":"e_1_2_2_95_1","volume-title":"DeepMapping: Learned Data Mapping for Lossless Compression and Efficient Lookup. In 2024 IEEE 40th International Conference on Data Engineering (ICDE). IEEE, 1-14","author":"Zhou Lixi","year":"2024","unstructured":"Lixi Zhou, K Sel\u00e7uk Candan, and Jia Zou. 2024a. DeepMapping: Learned Data Mapping for Lossless Compression and Efficient Lookup. In 2024 IEEE 40th International Conference on Data Engineering (ICDE). IEEE, 1-14."},{"key":"e_1_2_2_96_1","doi-asserted-by":"publisher","DOI":"10.14778\/3547305.3547325"},{"key":"e_1_2_2_97_1","first-page":"717","article-title":"Serving Deep Learning Models from Relational Databases","volume":"27","author":"Zhou Lixi","year":"2024","unstructured":"Lixi Zhou, Qi Lin, Kanchan Chowdhury, Saif Masood, Alexandre Eichenberger, Hong Min, Alexander Sim, Jie Wang, Yida Wang, Kesheng Wu, et al., 2024b. Serving Deep Learning Models from Relational Databases. Advances in Database Technology-EDBT, Vol. 27, 3 (2024), 717-724.","journal-title":"Advances in Database Technology-EDBT"},{"key":"e_1_2_2_98_1","doi-asserted-by":"publisher","DOI":"10.48786\/EDBT.2024.61"},{"key":"e_1_2_2_99_1","doi-asserted-by":"publisher","DOI":"10.14778\/3311880.3311885"},{"key":"e_1_2_2_100_1","doi-asserted-by":"publisher","DOI":"10.1007\/s00778-020-00605-w"}],"container-title":["Proceedings of the ACM on Management of Data"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/dl.acm.org\/doi\/pdf\/10.1145\/3725340","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,31]],"date-time":"2026-03-31T18:50:31Z","timestamp":1774983031000},"score":1,"resource":{"primary":{"URL":"https:\/\/dl.acm.org\/doi\/10.1145\/3725340"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,6,17]]},"references-count":100,"journal-issue":{"issue":"3","published-print":{"date-parts":[[2025,6,17]]}},"alternative-id":["10.1145\/3725340"],"URL":"https:\/\/doi.org\/10.1145\/3725340","relation":{},"ISSN":["2836-6573"],"issn-type":[{"value":"2836-6573","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,6,17]]}}}