AI & ML
Is 4-Bit Quantization More Than Noise to a Chat Model's Safety Behaviour? Attributing Behavioural Shifts with Matched Perturbations
arXiv cs.CR0 views
arXiv:2611.22186v1 Announce Type: new Abstract: Chat models on phones and laptops are usually 4-bit copies of the model that was evaluated, and audits of such copies disagree about what quantization does to refusal. A 4-bit build is one of many weight perturbations of the same size, so a change in its behaviour could come from its specific rounding error or from any perturbation that large. In a pre-registered study of three 3-4B chat models, with paired tests, blind human labels and an 8-bit null, we find two 4-bit shifts of opposite sign: with a compliant reply prefilled, Qwen2.5-3B continues into harmful content more often, and on held-out jailbreaks Llama-3.2-3B complies 10.3 points less often. Matched perturbations separate the two. Shuffled and Gaussian versions of the error average close to zero, while scaling the real error, from its mirror image to twice its size, changes behaviour near-linearly. At half the real error, Llama's likelihood is almost unchanged, yet it already complies 5.8 points less often, and llama.cpp's Q4_K_M, at the same bits per weight, moves neither model. Refusal under 4-bit quantization depends on the specific rounding error, not on bits per weight or likelihood, so each build has to be compared with its original.
Source:https://arxiv.org/abs/2611.22186PreviousHierarchical Personalized Federated Learning for Fine-Grained Intrusion Detection: A Multi-Class Analysis in Heterogeneous Attack EnvironmentsNextGapLeak: When Privacy-Preserving LLM-based Web Agents Become a Privacy Risk
Related posts
BenchAudit – Auditing LLM benchmarks for leaks, data reuse, and harness gaps
0Hacker News New
Guillotine DJ – Announcement Trailer
0Hacker News New
Researchers hijack Hugging Face account for popular Kokoro AI model
0Hacker News New
Automated LLM Fine-Tuning: GPU Selection, Dataset Analysis and Feedback
0Hacker News New
Human-written Launch HN posts get more upvotes than AI-generated ones
0Hacker News New

Parallels Desktop review
0TechRadar