How Fragile Is On-Device Language Model Safety? Localizing Safety-Critical Parameters for Sparse Fault Analysis

본문 미리보기

arXiv:2610.09000v1 Announce Type: new Abstract: As small language models (SLMs) are increasingly deployed on resource-constrained and on-device platforms, including as components of agentic systems, the integrity of locally stored model parameters becomes an important safety concern. We investigate whether safety-sensitive behavior in LLaMA-2-7B-Chat is concentrated within a sparse subset of parameters, creating a reduced fault surface for targeted analysis. We study two complementary localizat

How Fragile Is On-Device Language Model Safety? Localizing Safety-Critical Parameters for Sparse Fault Analysis

본문 미리보기

이 글이 만들어진 과정

관련 글

AudioShake Launches The Refinery to Turn Overlapping Conversations Into AI Training Data

Lava Finds Thousands of Exposed GPU Servers and a High-Severity NVIDIA Monitoring Flaw

Roundtables: A Conversation With the Creator of AI-Designed Viruses

ttok 0.4