mlsanityData to 7 Oct 2026

Investigating the Impact of Quantization Methods on the Safety and Reliability of Large Language Models

arXiv 2502.15799 ↗First posted 18 Feb 2025cs.CR cs.AIcited by 17

On the map: LLM reasoning → Reasoning for agents, planning and code

Abstract (arXiv)

Large Language Models (LLMs) are powerful tools for modern applications, but their computational demands limit accessibility. Quantization offers efficiency gains, yet its impact on safety and trustworthiness remains poorly understood. To address this, we introduce OpenMiniSafety, a human-curated safety dataset with 1.067 challenging questions to rigorously evaluate model behavior. We publicly release human safety evaluations for four LLMs (both quantized and full-precision), totaling 4.268 annotated question-answer pairs. By assessing 66 quantized variants of these models using four post-training quantization (PTQ) and two quantization-aware training (QAT) methods across four safety benchmarks including human-centric evaluations we uncover critical safety performance trade-offs. Our results show both PTQ and QAT can degrade safety alignment, with QAT techniques like QLORA or STE performing less safely. No single method consistently outperforms others across benchmarks, precision settings, or models, highlighting the need for safety-aware compression strategies. Furthermore, precision-specialized methods (e.g., QUIK and AWQ for 4-bit, AQLM and Q-PET for 2-bit) excel at their target precision, meaning that these methods are not better at compressing but rather different approaches.

Key facts. Cited by 17 papers in the map (data through 7 Oct 2026). Placed in Reasoning for agents, planning and code, one of 17 directions in LLM reasoning. Builds on 54 of its 76 references that are in the map. First posted on arXiv 18 Feb 2025.

Builds on

54 of 76 references.

References · 54

  1. 2022Training language models to follow instructions with human feedbackcited by 15,712
  2. 2024The Llama 3 Herd of Modelscited by 14,192
  3. 2023Direct Preference Optimization: Your Language Model is Secretly a Reward Modelcited by 7,916
  4. 2022Self-Consistency Improves Chain of Thought Reasoning in Language Modelscited by 5,601
  5. 2024GPT-4o System Cardcited by 4,467
  6. 2018HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answeringcited by 3,686
  7. 2023QLoRA: Efficient Finetuning of Quantized LLMscited by 3,355
  8. 2022Training Compute-Optimal Large Language Modelscited by 2,927
  9. 2013Estimating or Propagating Gradients Through Stochastic Neurons for Conditional Computationcited by 2,031
  10. 2024Gemma 2: Improving Open Language Models at a Practical Sizecited by 1,928
  11. 2024Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameterscited by 1,928
  12. 2022GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformerscited by 1,838
38 more
  1. 2022Language Models (Mostly) Know What They Knowcited by 1,783
  2. 2023Jailbroken: How Does LLM Safety Training Fail?cited by 1,565
  3. 2024HarmBench: A Standardized Evaluation Framework for Automated Red Teaming and Robust Refusalcited by 1,330
  4. 2023AWQ: Activation-aware Weight Quantization for LLM Compression and Accelerationcited by 1,234
  5. 2023Fine-tuning Aligned Language Models Compromises Safety, Even When Users Do Not Intend To!cited by 1,106
  6. 2024Qwen2.5-Math Technical Report: Toward Mathematical Expert Model via Self-Improvementcited by 1,057
  7. 2024Refusal in Language Models Is Mediated by a Single Directioncited by 960
  8. 2022Red Teaming Language Models to Reduce Harms: Methods, Scaling Behaviors, and Lessons Learnedcited by 702
  9. 2019Learned Step Size Quantizationcited by 604
  10. 2023QuIP: 2-Bit Quantization of Large Language Models With Guaranteescited by 397
  11. 2025Scaling up Test-Time Compute with Latent Reasoning: A Recurrent Depth Approachcited by 372
  12. 2022Optimal Brain Compression: A Framework for Accurate Post-Training Quantization and Pruningcited by 317
  13. 2023AlignScore: Evaluating Factual Consistency with a Unified Alignment Functioncited by 284
  14. 2024LLMs instead of Human Judges? A Large Scale Empirical Study across 20 NLP Evaluation Taskscited by 236
  15. 2024Does Fine-Tuning LLMs on New Knowledge Encourage Hallucinations?cited by 227
  16. 2024SALAD-Bench: A Hierarchical and Comprehensive Safety Benchmark for Large Language Modelscited by 225
  17. 2023ToxicChat: Unveiling Hidden Challenges of Toxicity Detection in Real-World User-AI Conversationcited by 224
  18. 2024Extreme Compression of Large Language Models via Additive Quantizationcited by 223
  19. 2024Assessing the Brittleness of Safety Alignment via Pruning and Low-Rank Modificationscited by 220
  20. 2024BiLLM: Pushing the Limit of Post-Training Quantization for LLMscited by 137
  21. 2023Safety Assessment of Chinese Large Language Modelscited by 82
  22. 2024Evaluating Quantized Large Language Modelscited by 75
  23. 2023SimpleSafetyTests: a Test Suite for Identifying Critical Safety Risks in Large Language Modelscited by 72
  24. 2024Can multiple-choice questions really be useful in detecting the abilities of LLMs?cited by 66
  25. 2024When Do LLMs Need Retrieval Augmentation? Mitigating LLMs' Overconfidence Helps Retrieval Augmentationcited by 59
  26. 2024Safetywashing: Do AI Safety Benchmarks Actually Measure Safety Progress?cited by 56
  27. 2024A Comprehensive Evaluation of Quantization Strategies for Large Language Modelscited by 47
  28. 2024FLAME: Factuality-Aware Alignment for Large Language Modelscited by 46
  29. 2022SafeText: A Benchmark for Exploring Physical Safety in Language Modelscited by 41
  30. 2024Accurate LoRA-Finetuning Quantization of LLMs via Information Retentioncited by 40
  31. 2023QUIK: Towards End-to-End 4-Bit Inference on Generative Large Language Modelscited by 39
  32. 2023Fake Alignment: Are LLMs Really Aligned Well?cited by 39
  33. 2024NLSR: Neuron-Level Safety Realignment of Large Language Models Against Harmful Fine-Tuningcited by 37
  34. 2025Adaptive Retrieval Without Self-Knowledge? Bringing Uncertainty Back Homecited by 33
  35. 2024Beyond Perplexity: Multi-dimensional Safety Evaluation of LLM Compressioncited by 32
  36. 2024Meta Reasoning for Large Language Modelscited by 29
  37. 2024Are Large Language Models More Honest in Their Probabilistic or Verbalized Confidence?cited by 23
  38. 2024LLMCBench: Benchmarking Large Language Model Compression for Efficient Deploymentcited by 16

50 most cited of 54 All 54 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 8; 2026: 9.

17 papers.

Sort:

By area:

  1. 2025Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Modelscited by 381
  2. 202570% Size, 100% Accuracy: Lossless LLM Compression for Efficient GPU Inference via Dynamic-Length Float (DFloat11)cited by 23
  3. 2025Low-Precision Training of Large Language Models: Methods, Challenges, and Opportunitiescited by 11
  4. 2026Quantization Undoes Alignment: Bias Emergence in Compressed LLMs Across Models and Precision Levelscited by 5
  5. 2025Geometric Metrics and LLMs: What They Measure and When They Workcited by 4
  6. 2025Safety-Preserving PTQ via Contrastive Alignment Losscited by 3
  7. 2026FLIPS: Instance-Fingerprinting for LLMs via Pseudo-random Sequencescited by 3
  8. 2026Float8@2bits: Entropy Coding Enables Data-Free Model Compressioncited by 2
  9. 2026Weight Pruning Amplifies Bias: A Multi-Method Study of Compressed LLMs for Edge AIcited by 2
  10. 2025From 2:4 to 8:16 sparsity patterns in LLMs for Outliers and Weights with Variance Correctioncited by 1
  11. 2026Quality Is Not a Safety Proxy Under Quantizationcited by 1
  12. 2025Motivating Next-Gen Accelerators with Flexible (N:M) Activation Sparsity via Benchmarking Lightweight Post-Training Sparsification Approachescited by 0
5 more
  1. 2025Enhancing Trustworthiness with Mixed Precision: Benchmarks, Opportunities, and Challengescited by 0
  2. 2026Preserving Fairness and Safety in Quantized LLMs Through Critical Weight Protectioncited by 0
  3. 2026The Joint Effect of Quantization and Sampling Temperature on LLM Safety Alignment: A Factorial Analysiscited by 0
  4. 2026On the Efficiency-Safety Dilemma in Large Reasoning Modelscited by 0
  5. 2026MOMAT: Mixture of Multiple Atlases for Low-Power Jailbreak Defense of Quantized LLMscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Safety Arithmetic: A Framework for Test-time Safety Alignment of Language Models by Steering Parameters and Activations
  2. 2024Navigating the Safety Landscape: Measuring Risks in Finetuning Large Language Models
  3. 2025CASE-Bench: Context-Aware SafEty Benchmark for Large Language Models
  4. 2026Q-realign: Piggybacking Realignment on Quantization for Safe and Efficient LLM Deployment
  5. 2026SafeSci: Safety Evaluation of Large Language Models in Science Domains and Beyond
  6. 2024LongSafety: Enhance Safety for Long-Context LLMs
  7. 2023All Languages Matter: On the Multilingual Safety of Large Language Models
  8. 2025Activation Approximations Can Incur Safety Vulnerabilities Even in Aligned LLMs: Comprehensive Analysis and Defense