Panacea: Mitigating harmful fine-tuning for large language models via post-fine-tuning perturbation

Research output: Journal article publication › Journal article › Academic research › peer-review

Original language	English
Pages (from-to)	169951-169985
Number of pages	35
Journal	Advances in Neural Information Processing Systems
Volume	38
Publication status	Published - 2026

Cite this

@article{fc6b610bb226401d855c230bb806d430,

title = "Panacea: Mitigating harmful fine-tuning for large language models via post-fine-tuning perturbation",

author = "Yibo Wang and Tiansheng Huang and Li Shen and Huanjin Yao and Haotian Luo and Rui Liu and Naiqiang Tan and Jiaxing Huang and Dacheng Tao",

year = "2026",

language = "English",

volume = "38",

pages = "169951--169985",

journal = "Advances in Neural Information Processing Systems",

issn = "1049-5258",

publisher = "Neural information processing systems foundation",

}

TY - JOUR

T1 - Panacea: Mitigating harmful fine-tuning for large language models via post-fine-tuning perturbation

AU - Wang, Yibo

AU - Huang, Tiansheng

AU - Shen, Li

AU - Yao, Huanjin

AU - Luo, Haotian

AU - Liu, Rui

AU - Tan, Naiqiang

AU - Huang, Jiaxing

AU - Tao, Dacheng

PY - 2026

Y1 - 2026

M3 - Journal article

SN - 1049-5258

VL - 38

SP - 169951

EP - 169985

JO - Advances in Neural Information Processing Systems

JF - Advances in Neural Information Processing Systems

ER -