mlsanityData to 7 Oct 2026

Adversarial Contrastive Learning for LLM Quantization Attacks

arXiv 2601.02680 ↗First posted 6 Jan 2026cs.CR cs.LGcited by 4

On the map: LLM safety and security → Jailbreaking

Abstract (arXiv)

Model quantization is critical for deploying large language models (LLMs) on resource-constrained hardware, yet recent work has revealed severe security risks that benign LLMs in full precision may exhibit malicious behaviors after quantization. In this paper, we propose Adversarial Contrastive Learning (ACL), a novel gradient-based quantization attack that achieves superior attack effectiveness by explicitly maximizing the gap between benign and harmful responses probabilities. ACL formulates the attack objective as a triplet-based contrastive loss, and integrates it with a projected gradient descent two-stage distributed fine-tuning strategy to ensure stable and efficient optimization. Extensive experiments demonstrate ACL's remarkable effectiveness, achieving attack success rates of 86.00% for over-refusal, 97.69% for jailbreak, and 92.40% for advertisement injection, substantially outperforming state-of-the-art methods by up to 44.67%, 18.84%, and 50.80%, respectively.

Key facts. Cited by 4 papers in the map (data through 7 Oct 2026). Placed in Jailbreaking, one of 12 directions in LLM safety and security. Builds on 21 of its 42 references that are in the map. First posted on arXiv 6 Jan 2026.

Builds on

21 of 42 references.

Closest ancestors · 2

Same direction, newest first

  1. 2024Latent Adversarial Training Improves Robustness to Persistent Harmful Behaviors in LLMs22 Jul
  2. 2023Universal and Transferable Adversarial Attacks on Aligned Language Models27 Jul

Other references · 19

  1. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  2. 2019Sentence-BERT: Sentence Embeddings using Siamese BERT-Networkscited by 9,049
  3. 2020Measuring Massive Multitask Language Understandingcited by 7,768
  4. 2013Efficient Estimation of Word Representations in Vector Spacecited by 4,346
  5. 2023QLoRA: Efficient Finetuning of Quantized LLMscited by 3,355
  6. 2015FaceNet: A Unified Embedding for Face Recognition and Clusteringcited by 2,833
  7. 2022GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformerscited by 1,838
  8. 2023AWQ: Activation-aware Weight Quantization for LLM Compression and Accelerationcited by 1,234
  9. 2020Keystroke Dynamics as Part of Lifeloggingcited by 334
  10. 2024Improving Alignment and Robustness with Circuit Breakerscited by 302
  11. 2023Poisoning Web-Scale Training Datasets is Practicalcited by 276
  12. 2019A Study of BFLOAT16 for Deep Learning Trainingcited by 227
7 more
  1. 2024An empirical study of LLaMA3 quantization: from LLMs to MLLMscited by 63
  2. 2025Representation Bending for Large Language Model Safetycited by 28
  3. 2024Distinguishing LLM-generated from Human-written Code by Contrastive Learningcited by 16
  4. 2024DecDEC: A Systems Approach to Advancing Low-Bit LLM Quantizationcited by 7
  5. 2025Improving Large Language Model Safety with Contrastive Representation Learningcited by 7
  6. 2025Fewer Weights, More Problems: A Practical Attack on LLM Pruningcited by 6
  7. 2025Supervised Fine-Tuning or Contrastive Learning? Towards Better Multimodal LLM Rerankingcited by 6

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 4.

4 papers.

  1. 2026Hair-Trigger Alignment: Black-Box Evaluation Cannot Guarantee Post-Update Alignmentcited by 3
  2. 2026Trusted Weights, Treacherous Optimizations? Optimization-Triggered Backdoor Attacks on LLMscited by 1
  3. 2026Widening the Gap: Exploiting LLM Quantization via Outlier Injectioncited by 0
  4. 2026AGENTQ: Quantization-Conditioned Backdoor Attacks on LLM Agentscited by 0