{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,2]],"date-time":"2026-07-02T07:02:18Z","timestamp":1782975738461,"version":"3.54.5"},"reference-count":90,"publisher":"IEEE","license":[{"start":{"date-parts":[[2026,5,18]],"date-time":"2026-05-18T00:00:00Z","timestamp":1779062400000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-029"},{"start":{"date-parts":[[2026,5,18]],"date-time":"2026-05-18T00:00:00Z","timestamp":1779062400000},"content-version":"stm-asf","delay-in-days":0,"URL":"https:\/\/doi.org\/10.15223\/policy-037"}],"funder":[{"DOI":"10.13039\/501100001381","name":"National Research Foundation, Singapore","doi-asserted-by":"publisher","id":[{"id":"10.13039\/501100001381","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"published-print":{"date-parts":[[2026,5,18]]},"DOI":"10.1109\/sp63933.2026.00072","type":"proceedings-article","created":{"date-parts":[[2026,7,1]],"date-time":"2026-07-01T19:34:20Z","timestamp":1782934460000},"page":"3358-3376","source":"Crossref","is-referenced-by-count":0,"title":["ENCHTABLE: Unified Safety Alignment Transfer in Fine-Tuned Large Language Models"],"prefix":"10.1109","author":[{"given":"Jialin","family":"Wu","sequence":"first","affiliation":[{"name":"Ant Group,China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Kecen","family":"Li","sequence":"additional","affiliation":[{"name":"Ant Group,China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Zhicong","family":"Huang","sequence":"additional","affiliation":[{"name":"Ant Group,China"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xinfeng","family":"Li","sequence":"additional","affiliation":[{"name":"Nanyang Technological University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Xiaofeng","family":"Wang","sequence":"additional","affiliation":[{"name":"Nanyang Technological University"}],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Cheng","family":"Hong","sequence":"additional","affiliation":[{"name":"Ant Group,China"}],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"263","reference":[{"key":"ref1","article-title":"Llama2: Open foundation and fine-tuned chat models","author":"Touvron","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref2","article-title":"The llama 3 herd of models","author":"Grattafiori","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref3","doi-asserted-by":"publisher","DOI":"10.52202\/068431-2011"},{"key":"ref4","article-title":"Gpt-4 technical report","author":"Achiam","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref5","article-title":"Code llama: Open foundation models for code","author":"Roziere","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref6","article-title":"Qwen2. 5-coder technical report","author":"Hui","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref7","article-title":"Biomedgpt: Open multimodal generative pre-trained transformer for biomedicine","author":"Luo","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref8","article-title":"Qwen2. 5 technical report","author":"Yang","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref9","article-title":"Metamath: Bootstrap your own mathematical questions for large language models","volume-title":"The Twelfth International Conference on Learning Representations, ICLR2024","author":"Yu","year":"2024"},{"key":"ref10","article-title":"Fine-tuning aligned language models compromises safety, even when users do not intend to!","author":"Qi","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref11","doi-asserted-by":"publisher","DOI":"10.1109\/tpami.2026.3660147"},{"key":"ref12","article-title":"Safe rlhf: Safe reinforcement learning from human feedback","author":"Dai","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref13","article-title":"Safety-tuned llamas: Lessons from improving the safety of large language models that follow instructions","author":"Bianchi","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref14","article-title":"Editing models with task arithmetic","volume-title":"The Eleventh International Conference on Learning Representations, ICLR2023","author":"Ilharco","year":"2023"},{"key":"ref15","article-title":"Mastering task arithmetic:$\\tau$ jp as a key indicator for weight disentanglement","volume-title":"The Thirteenth International Conference on Learning Representations, ICLR 2025","author":"Yoshida","year":"2025"},{"key":"ref16","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2913"},{"key":"ref17","article-title":"Mistral 7b","volume":"abs\/2310.06825","author":"Jiang","year":"2023","journal-title":"CoRR"},{"key":"ref18","doi-asserted-by":"publisher","DOI":"10.52202\/079017-1790"},{"key":"ref19","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2021.emnlp-main.446"},{"key":"ref20","article-title":"A survey of large language models","author":"Zhao","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref21","article-title":"Openai o1 system card","author":"Jaech","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref22","article-title":"Qwen technical report","author":"Bai","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref23","article-title":"Attention is all you need","volume-title":"Conference on Neural Information Processing Systems","author":"Vaswani"},{"key":"ref24","article-title":"A survey of transformers","author":"Lin","year":"2021","journal-title":"arXiv preprint arXiv"},{"key":"ref25","article-title":"Physics of language models: Part 3.1, knowledge storage and extraction","author":"Allen-Zhu","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref26","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.emnlp-demos.2"},{"key":"ref27","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.emnlp-main.751"},{"key":"ref28","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.acl-long.581"},{"key":"ref29","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.171"},{"key":"ref30","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2020.acl-demos.22"},{"key":"ref31","doi-asserted-by":"publisher","DOI":"10.52202\/079017-2163"},{"key":"ref32","article-title":"Editable neural networks","volume-title":"8th International Conference on Learning Representations, ICLR2020","author":"Sinitsin","year":"2020"},{"key":"ref33","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-demos.9"},{"key":"ref34","doi-asserted-by":"publisher","DOI":"10.1145\/3698590"},{"key":"ref35","doi-asserted-by":"publisher","DOI":"10.52202\/068431-1262"},{"key":"ref36","article-title":"Fast model editing at scale","volume-title":"The Tenth International Conference on Learning Representations, ICLR2022, Virtual Event","author":"Mitchell","year":"2022"},{"key":"ref37","article-title":"Transformer-patcher: One mistake worth one neuron","volume-title":"The Eleventh International Conference on Learning Representations, ICLR2023","author":"Huang","year":"2023"},{"key":"ref38","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2022.findings-emnlp.438"},{"key":"ref39","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2079"},{"key":"ref40","article-title":"Neural tangent kernel: Convergence and generalization in neural networks","volume-title":"Advances in Neural Information Processing Systems31: Annual Conference on Neural Information Processing Systems 2018, NeurIPS 2018","author":"Jacot","year":"2018"},{"key":"ref41","article-title":"Parameter efficient multi-task model fusion with partial linearization","author":"Tang","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref42","article-title":"Lora: Low-rank adaptation of large language models","volume-title":"The Tenth International Conference on Learning Representations, ICLR2022, Virtual Event","author":"Hu","year":"2022"},{"key":"ref43","article-title":"Task singular vectors: Reducing task interference in model merging","author":"Gargiulo","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref44","article-title":"Safety layers in aligned large language models: The key to lm security","author":"Li","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref45","article-title":"Lora: Low-rank adaptation of large language models","volume-title":"The Tenth International Conference on Learning Representations, ICLR2022, Virtual Event","author":"Hu","year":"2022"},{"key":"ref46","article-title":"PyTorch: An imperative style, high-performance deep learning library","volume-title":"Conference on Neural Information Processing Systems","author":"Paszke"},{"key":"ref47","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-demos.38"},{"key":"ref48","article-title":"Code-llama-3\u20138b","volume-title":"ajibawa 2023","year":"2024"},{"key":"ref49","doi-asserted-by":"publisher","DOI":"10.52202\/079017-2877"},{"key":"ref50","volume-title":"Bio-medical: A high-performance biomedical language model","year":"2024"},{"key":"ref51","volume-title":"Openhermes2.5-mistral-7b","year":"2024"},{"key":"ref52","article-title":"Code-290k-sharegpt","volume-title":"ajibawa 2023","year":"2024"},{"key":"ref53","article-title":"Orca-math: Unlocking the potential of slms in grade school math","author":"Mitra","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref54","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-acl.762"},{"key":"ref55","doi-asserted-by":"publisher","DOI":"10.52202\/079017-2877"},{"key":"ref56","doi-asserted-by":"publisher","DOI":"10.52202\/079017-2078"},{"key":"ref57","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.762"},{"key":"ref58","doi-asserted-by":"publisher","DOI":"10.52202\/075280-0310"},{"key":"ref59","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-73226-3_16"},{"key":"ref60","doi-asserted-by":"publisher","DOI":"10.1007\/978-3-031-72784-9_12"},{"key":"ref61","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.303"},{"key":"ref62","doi-asserted-by":"publisher","DOI":"10.1038\/s42256-023-00765-8"},{"key":"ref63","article-title":"Sorry-bench: Systematically evaluating large language model safety refusal","volume-title":"The Thirteenth International Conference on Learning Representations, ICLR2025","author":"Xie","year":"2025"},{"key":"ref64","doi-asserted-by":"publisher","DOI":"10.52202\/075280-0943"},{"key":"ref65","volume-title":"Llm math evaluation harness","author":"Gou","year":"2024"},{"key":"ref66","article-title":"Lessons from the trenches on reproducible evaluation of language models","author":"Biderman","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref67","doi-asserted-by":"publisher","DOI":"10.52202\/075280-1072"},{"key":"ref68","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.findings-acl.235"},{"key":"ref69","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2023.acl-long.244"},{"key":"ref70","article-title":"Language model unalignment: Parametric red-teaming to expose hidden harms and biases","author":"Bhardwaj","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref71","article-title":"Evaluating large language models trained on code","author":"Chen","year":"2021","journal-title":"arXiv preprint arXiv"},{"key":"ref72","article-title":"Training verifiers to solve math word problems","author":"Cobbe","year":"2021","journal-title":"arXiv preprint arXiv"},{"key":"ref73","article-title":"Measuring mathematical problem solving with the MATH dataset","volume-title":"Proceedings of the Neural Information Processing Systems Track on Datasets and Benchmarks 1, NeurIPS Datasets and Benchmarks2021","author":"Hendrycks","year":"2021"},{"key":"ref74","doi-asserted-by":"publisher","DOI":"10.3390\/app11146421"},{"key":"ref75","article-title":"Harmbench: A standardized evaluation framework for automated red teaming and robust refusal","author":"Mazeika","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref76","article-title":"Making them ask and answer: Jailbreaking large language models in few queries via disguise and reconstruction","volume-title":"33rd USENIX Security Symposium, USENIX Security2024","author":"Liu","year":"2024"},{"key":"ref77","article-title":"A general language assistant as a laboratory for alignment","author":"Askell","year":"2021","journal-title":"arXiv preprint arXiv"},{"key":"ref78","doi-asserted-by":"publisher","DOI":"10.18653\/v1\/2024.acl-long.481"},{"key":"ref79","doi-asserted-by":"publisher","DOI":"10.52202\/075280-1797"},{"key":"ref80","article-title":"Training a helpful and harmless assistant with reinforcement learning from human feedback","author":"Bai","year":"2022","journal-title":"arXiv preprint arXiv"},{"key":"ref81","article-title":"Fundamental limitations of alignment in large language models","author":"Wolf","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref82","doi-asserted-by":"publisher","DOI":"10.52202\/079017-2356"},{"key":"ref83","doi-asserted-by":"publisher","DOI":"10.52202\/079017-0402"},{"key":"ref84","article-title":"Fine-tuning can cripple your foundation model; preserving features may be the solution","author":"Mukhoti","year":"2023","journal-title":"arXiv preprint arXiv"},{"key":"ref85","article-title":"Seal: Safety-enhanced aligned 11 m fine-tuning via bilevel data selection","author":"Shen","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref86","doi-asserted-by":"publisher","DOI":"10.1016\/j.knosys.2024.112701"},{"key":"ref87","article-title":"Antidote: Post-fine-tuning safety alignment for large language models against harmful fine-tuning","author":"Huang","year":"2024","journal-title":"arXiv preprint arXiv"},{"key":"ref88","article-title":"Deepseek-r1-distill-qwen-7b-japanese","volume-title":"lightblue","year":"2025"},{"key":"ref89","doi-asserted-by":"publisher","DOI":"10.52202\/075280-2338"},{"key":"ref90","article-title":"Aisafetylab: A comprehensive framework for ai safety evaluation and improvement","author":"Zhang","year":"2025","journal-title":"arXiv preprint arXiv"}],"event":{"name":"2026 IEEE Symposium on Security and Privacy (SP)","location":"San Francisco, CA, USA","start":{"date-parts":[[2026,5,18]]},"end":{"date-parts":[[2026,5,21]]}},"container-title":["2026 IEEE Symposium on Security and Privacy (SP)"],"original-title":[],"link":[{"URL":"http:\/\/xplorestaging.ieee.org\/ielx8\/11573355\/11573356\/11573502.pdf?arnumber=11573502","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,2]],"date-time":"2026-07-02T05:30:33Z","timestamp":1782970233000},"score":1,"resource":{"primary":{"URL":"https:\/\/ieeexplore.ieee.org\/document\/11573502\/"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,5,18]]},"references-count":90,"URL":"https:\/\/doi.org\/10.1109\/sp63933.2026.00072","relation":{},"subject":[],"published":{"date-parts":[[2026,5,18]]}}}