{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,4,28]],"date-time":"2026-04-28T05:40:42Z","timestamp":1777354842278,"version":"3.51.4"},"reference-count":57,"publisher":"Institute of Electrical and Electronics Engineers (IEEE)","license":[{"start":{"date-parts":[[2026,1,1]],"date-time":"2026-01-01T00:00:00Z","timestamp":1767225600000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/legalcode"}],"funder":[{"name":"National Key Research and Development Program of China","award":["2021YFB3100300"],"award-info":[{"award-number":["2021YFB3100300"]}]},{"DOI":"10.13039\/501100001809","name":"Nature Science Foundation of China","doi-asserted-by":"publisher","award":["U20A20178"],"award-info":[{"award-number":["U20A20178"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100001809","name":"Nature Science Foundation of China","doi-asserted-by":"publisher","award":["62072395"],"award-info":[{"award-number":["62072395"]}],"id":[{"id":"10.13039\/501100001809","id-type":"DOI","asserted-by":"publisher"}]},{"name":"\u201cPioneer\u201d and \u201cLeading Goose\u201d Research and Development Program of Zhejiang Province","award":["2024C01169"],"award-info":[{"award-number":["2024C01169"]}]},{"name":"Nature Science Foundation of China","award":["62206207"],"award-info":[{"award-number":["62206207"]}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["IEEE Trans.Inform.Forensic Secur."],"published-print":{"date-parts":[[2026]]},"DOI":"10.1109\/tifs.2026.3679194","type":"journal-article","created":{"date-parts":[[2026,3,30]],"date-time":"2026-03-30T20:05:29Z","timestamp":1774901129000},"page":"4163-4175","source":"Crossref","is-referenced-by-count":0,"title":["Differentially Private Zeroth-Order Methods for Scalable Large Language Model Fine-Tuning"],"prefix":"10.1109","volume":"21","author":[{"ORCID":"https:\/\/orcid.org\/0009-0005-5828-5437","authenticated-orcid":false,"given":"Zhihao","family":"Liu","sequence":"first","affiliation":[{"name":"State Key Laboratory of Blockchain and Data Security, Zhejiang University, Hangzhou, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-4110-2068","authenticated-orcid":false,"given":"Jian","family":"Lou","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Blockchain and Data Security, Zhejiang University, Hangzhou, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Wenjie","family":"Bao","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Blockchain and Data Security, Zhejiang University, Hangzhou, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-5780-6898","authenticated-orcid":false,"given":"Yuke","family":"Hu","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Blockchain and Data Security, Zhejiang University, Hangzhou, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Wenli","family":"Wang","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Blockchain and Data Security, Zhejiang University, Hangzhou, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0001-7872-6969","authenticated-orcid":false,"given":"Zhan","family":"Qin","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Blockchain and Data Security, Zhejiang University, Hangzhou, China"}],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0000-0002-1969-2591","authenticated-orcid":false,"given":"Kui","family":"Ren","sequence":"additional","affiliation":[{"name":"State Key Laboratory of Blockchain and Data Security, Zhejiang University, Hangzhou, China"}],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"263","reference":[{"key":"ref1","article-title":"OPT: Open pre-trained transformer language models","author":"Zhang","year":"2022","journal-title":"arXiv:2205.01068"},{"key":"ref2","article-title":"LLaMA: Open and efficient foundation language models","author":"Touvron","year":"2023","journal-title":"arXiv:2302.13971"},{"issue":"8","key":"ref3","first-page":"9","article-title":"Language models are unsupervised multitask learners","volume":"1","author":"Radford","year":"2019","journal-title":"OpenAI Blog"},{"key":"ref4","article-title":"Improving language understanding by generative pre-training","volume-title":"Proc. 32nd Conf. Neural Inf. Process. Syst.","author":"Radford","year":"2018"},{"key":"ref5","article-title":"LLaMA-adapter: Efficient fine-tuning of language models with zero-init attention","author":"Zhang","year":"2023","journal-title":"arXiv:2303.16199"},{"key":"ref6","article-title":"Fine-tuning language models from human preferences","author":"Ziegler","year":"2019","journal-title":"arXiv:1909.08593"},{"key":"ref7","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-030-32381-3_16"},{"key":"ref8","first-page":"267","article-title":"The secret sharer: Evaluating and testing unintended memorization in neural networks","volume-title":"Proc. 28th USENIX Secur. Symp.","author":"Carlini"},{"key":"ref9","first-page":"2633","article-title":"Extracting training data from large language models","volume-title":"Proc. 30th USENIX Secur. Symp.","author":"Carlini"},{"key":"ref10","doi-asserted-by":"publisher","DOI":"10.1007\/11787006_1"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.29012\/jpc.880"},{"key":"ref12","article-title":"Large language models can be strong differentially private learners","author":"Li","year":"2021","journal-title":"arXiv:2110.05679"},{"key":"ref13","article-title":"Exploring the limits of differentially private deep learning with group-wise clipping","author":"He","year":"2022","journal-title":"arXiv:2212.01539"},{"key":"ref14","doi-asserted-by":"publisher","DOI":"10.1145\/2976749.2978318"},{"key":"ref15","doi-asserted-by":"publisher","DOI":"10.1145\/3469877.3490594"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.52202\/075280-1808"},{"key":"ref17","doi-asserted-by":"publisher","DOI":"10.1109\/TC.2023.3236868"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2308"},{"key":"ref19","article-title":"Second-order fine-tuning without pain for LLMs: A Hessian informed zeroth-order optimizer","author":"Zhao","year":"2024","journal-title":"arXiv:2402.15173"},{"key":"ref20","article-title":"Revisiting zeroth-order optimization for memory-efficient LLM fine-tuning: A benchmark","author":"Zhang","year":"2024","journal-title":"arXiv:2402.11592"},{"key":"ref21","article-title":"Privacy-preserving prompt tuning for large language model services","author":"Li","year":"2023","journal-title":"arXiv:2305.06212"},{"key":"ref22","doi-asserted-by":"publisher","DOI":"10.52202\/075280-3358"},{"key":"ref23","first-page":"12208","article-title":"Large scale private learning via low-rank reparametrization","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Yu"},{"key":"ref24","first-page":"59210","article-title":"DPZERO: Private fine-tuning of language models without backpropagation","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Zhang"},{"key":"ref25","article-title":"Private fine-tuning of large language models with zeroth-order optimization","author":"Tang","year":"2024","journal-title":"arXiv:2401.04343"},{"key":"ref26","article-title":"Private zeroth-order nonsmooth nonconvex optimization","author":"Zhang","year":"2024","journal-title":"arXiv:2406.19579"},{"key":"ref27","article-title":"Differential privacy meets neural network pruning","author":"Adamczewski","year":"2023","journal-title":"arXiv:2303.04612"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.1109\/CVPR46437.2021.00502"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.52202\/068431-2137"},{"key":"ref30","article-title":"Pre-pruning and gradient-dropping improve differentially private image classification","author":"Adamczewski","year":"2023","journal-title":"arXiv:2306.11754"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.1007\/11681878_14"},{"key":"ref32","doi-asserted-by":"publisher","DOI":"10.1109\/9.119632"},{"key":"ref33","first-page":"2608","article-title":"Stagewise training accelerates convergence of testing error over sgd","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"32","author":"Yuan"},{"key":"ref34","first-page":"9317","article-title":"Differentially private stochastic optimization: New results in convex and non-convex settings","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"34","author":"Bassily"},{"key":"ref35","first-page":"30842","article-title":"Zeroth-order methods for constrained nonconvex nonsmooth stochastic optimization","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Liu"},{"key":"ref36","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-319-46128-1_50"},{"key":"ref37","article-title":"On exponential convergence of SGD in non-convex over-parametrized learning","author":"Bassily","year":"2018","journal-title":"arXiv:1811.02564"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.1007\/s10208-015-9296-2"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.1137\/120880811"},{"key":"ref40","first-page":"19935","article-title":"Gradient-free method for heavily constrained nonconvex optimization","volume-title":"Proc. Int. Conf. Mach. Learn.","author":"Shi"},{"key":"ref41","first-page":"6280","article-title":"Privacy amplification by subsampling: Tight analyses via couplings and divergences","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"31","author":"Balle"},{"key":"ref42","doi-asserted-by":"publisher","DOI":"10.48550\/ARXIV.1907.11692"},{"key":"ref43","article-title":"Making pre-trained language models better few-shot learners","author":"Gao","year":"2020","journal-title":"arXiv:2012.15723"},{"key":"ref44","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D13-1170"},{"key":"ref45","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D15-1075"},{"key":"ref46","article-title":"A broad-coverage challenge corpus for sentence understanding through inference","author":"Williams","year":"2017","journal-title":"arXiv:1704.05426"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.1145\/345508.345577"},{"key":"ref48","first-page":"3266","article-title":"Superglue: A stickier benchmark for general-purpose language understanding systems","volume-title":"Proc. Adv. Neural Inf. Process. Syst.","volume":"32","author":"Wang"},{"issue":"2","key":"ref49","first-page":"107","article-title":"The commitmentbank: Investigating projection in naturally occurring discourse","volume-title":"Proc. Sinn und Bedeutung","volume":"23","author":"De Marneffe"},{"key":"ref50","article-title":"BoolQ: Exploring the surprising difficulty of natural Yes\/No questions","author":"Clark","year":"2019","journal-title":"arXiv:1905.10044"},{"key":"ref51","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/N18-1023"},{"key":"ref52","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D16-1264"},{"key":"ref53","first-page":"79","article-title":"Europarl: A parallel corpus for statistical machine translation","volume-title":"Proc. Mach. Transl. Summit X, Papers","author":"Koehn"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2020.emnlp-main.213"},{"key":"ref55","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/K16-1028"},{"key":"ref56","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/D18-1206"},{"key":"ref57","first-page":"74","article-title":"Rouge: A package for automatic evaluation of summaries","volume-title":"Proc. Text Summarization Branches Out","author":"Lin"}],"container-title":["IEEE Transactions on Information Forensics and Security"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/10206\/11313711\/11457969.pdf?arnumber=11457969","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,4,28]],"date-time":"2026-04-28T04:48:45Z","timestamp":1777351725000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11457969\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026]]},"references-count":57,"URL":"https:\/\/doi.org\/10.1109\/tifs.2026.3679194","relation":{},"ISSN":["1556-6013","1556-6021"],"issn-type":[{"value":"1556-6013","type":"print"},{"value":"1556-6021","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026]]}}}