cs.LGNov 16, 2026

Model Immunization Beyond Condition Numbers: The Importance of Plateau Regions

Authors: Quoc Nguyen, Trung Le, Dinh Phung, Mehrtash Harandi

Organizations: Department of Electrical and Computer Systems Engineering, Monash University, Australia · Department of Data Science and AI, Monash University, Australia

Abstract

Harmful fine-tuning can rapidly compromise safety alignment in Large Language Models, underscoring the vulnerability of deployed models to post-release modification. While prior work motivates model immunization by inducing ill-conditioned curvature to slow harmful optimization, this approach is not scalable due to the high computational cost of Hessian estimation. We refine this paradigm by extending the theoretical framework to prioritize plateau regions alongside curvature. We further introduce an immunization algorithm based on efficient Hessian estimation and bounds on the extremal singular values of the Hessian. We demonstrate the proposed immunization framework in three applications: (i) Fine-tuning-as-a-service settings where benign user data may be mixed with poisoned harmful data, (ii) open-weight settings where an adversary can fine-tune on harmful-only data with varied training dynamics, and (iii) robust unlearning, where immunized models better resist relearning attacks after unlearning. Extensive experiments confirm that our approach achieves robust defense against adversarial fine-tuning while largely maintaining general utility and downstream trainability. Our code is available at https://github.com/minhquoc0712/BECON.

Figures & tables

Appendix figures & tables7 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. Antibody: Strengthening Defense Against Harmful Fine-Tuning for Large Language Models via Attenuating Harmful Gradient Influence

    Feb 28, 2026Quoc Minh Nguyen, Trung Le, Jing Wu +2Safety-Preserving Fine-Tuning of Language ModelsLLM Fine-Tuning

  2. Gradient Immunity: Null-Space Resistance to Malicious Fine-Tuning

    Aug 5, 2026Yuxuan Huang, Xingyu Zeng, Tianhang Zheng +1LLM Safety AlignmentLLM Security

  3. A Few Steps Further: Why Defenses Against Malicious Finetuning Erode Under Continued Training

    May 14, 2026Itay Zloczower, Eyal Lenga, Gilad Gressel +1Language Model Safety EvaluationAdversarial Attacks on LLMs