mlsanityData to 14 Oct 2026

Model Immunization from a Condition Number Perspective

arXiv 2505.23760 ↗First posted 29 May 2025cs.LGcited by 8

Not on the map. Why?

Abstract (arXiv)

Model immunization aims to pre-train models that are difficult to fine-tune on harmful tasks while retaining their utility on other non-harmful tasks. Though prior work has shown empirical evidence for immunizing text-to-image models, the key understanding of when immunization is possible and a precise definition of an immunized model remain unclear. In this work, we propose a framework, based on the condition number of a Hessian matrix, to analyze model immunization for linear models. Building on this framework, we design an algorithm with regularization terms to control the resulting condition numbers after pre-training. Empirical results on linear models and non-linear deep-nets demonstrate the effectiveness of the proposed algorithm on model immunization. The code is available at https://github.com/amberyzheng/model-immunization-cond-num.

Key facts. Cited by 8 papers in the map (data through 14 Oct 2026). Builds on 35 of its 62 references that are in the map. First posted on arXiv 29 May 2025.

Builds on

35 of 62 references.

References · 35

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,744
  3. 2020An Image is Worth 16x16 Words: Transformers for Image Recognition at Scalecited by 30,186
  4. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  5. 2021Convex optimizationcited by 2,515
  6. 2013Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computationcited by 2,079
  7. 2019A Comprehensive Survey on Transfer Learningcited by 1,160
  8. 2014Convex Optimization: Algorithms and Complexitycited by 575
  9. 2023Erasing Concepts from Diffusion Modelscited by 539
  10. 2019Certified Data Removal from Machine Learning Modelscited by 535
  11. 2022Safe Latent Diffusion: Mitigating Inappropriate Degeneration in Diffusion Modelscited by 513
  12. 2023Unified Concept Editing in Diffusion Modelscited by 388
23 more
  1. 2021Remember What You Want to Forget: Algorithms for Machine Unlearningcited by 371
  2. 2023Ablating Concepts in Text-to-Image Diffusion Modelscited by 317
  3. 2020Descent-to-Delete: Gradient-Based Methods for Machine Unlearningcited by 311
  4. 2023Forget-Me-Not: Learning to Forget in Text-to-Image Diffusion Modelscited by 301
  5. 2023Removing RLHF Protections in GPT-4 via Fine-Tuningcited by 171
  6. 2018The Malicious Use of Artificial Intelligence: Forecasting, Prevention, and Mitigationcited by 165
  7. 2020Variational Bayesian Unlearningcited by 117
  8. 2017Sticking the Landing: Simple, Lower-Variance Gradient Estimators for Variational Inferencecited by 98
  9. 2024Harmful Fine-tuning Attacks and Defenses for Large Language Models: A Surveycited by 90
  10. 2022PUMA: Performance Unchanged Model Augmentation for Training Data Removalcited by 58
  11. 2025Unlearning Isn't Deletion: Investigating Reversibility of Machine Unlearning in LLMscited by 42
  12. 2024Reconstruction Attacks on Machine Unlearning: Simple Models are Vulnerablecited by 25
  13. 2025Antidistillation Samplingcited by 25
  14. 2024Generative AI Misuse: A Taxonomy of Tactics and Insights from Real-World Datacited by 20
  15. 2023IMMA: Immunizing text-to-image Models against Malicious Adaptationcited by 18
  16. 2020A Study of Condition Numbers for First-Order Optimizationcited by 10
  17. 2020Fast and Near-Optimal Diagonal Preconditioningcited by 9
  18. 2024Weight Conditioning for Smooth Optimization of Neural Networkscited by 9
  19. 2023Structured Semidefinite Programming for Recovering Structured Preconditionerscited by 5
  20. 2024Learning to Obstruct Few-Shot Image Classification over Restricted Classescited by 5
  21. 2024(Almost) Smooth Sailing: Towards Numerical Stability of Neural Networks Through Differentiable Regularization of the Condition Numbercited by 5
  22. 2024Multi-concept Model Immunization through Differentiable Model Mergingcited by 4
  23. 2024Probe-Me-Not: Protecting Pre-trained Encoders from Malicious Probingcited by 2

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2025: 2; 2026: 6.

8 papers.

  1. 2025Gradient Surgery for Safe LLM Fine-Tuningcited by 11
  2. 2025Knowledge Distillation Detection for Open-weights Modelscited by 4
  3. 2026Surgery: Mitigating Harmful Fine-Tuning for Large Language Models via Attention Sinkcited by 4
  4. 2026Safeguarding LLM Fine-tuning via Push-Pull Distributional Alignmentcited by 2
  5. 2026Designing to Forget: Deep Semi-parametric Models for Unlearningcited by 1
  6. 2026Open-Weight LLM Fine-Tuning Defenses are Susceptible to Simple Attackscited by 1
  7. 2026Locking Pretrained Weights via Deep Low-Rank Residual Distillationcited by 0
  8. 2026The Distillation Game: Adaptive Evaluations & Efficient Defensescited by 0