惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

MyScale Blog
MyScale Blog
Jina AI
Jina AI
OSCHINA 社区最新新闻
OSCHINA 社区最新新闻
阮一峰的网络日志
阮一峰的网络日志
P
Proofpoint News Feed
Last Week in AI
Last Week in AI
博客园 - 司徒正美
Martin Fowler
Martin Fowler
T
Tailwind CSS Blog
B
Blog RSS Feed
Vercel News
Vercel News
博客园 - 聂微东
I
InfoQ
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
M
MIT News - Artificial intelligence
Recent Announcements
Recent Announcements
GbyAI
GbyAI
L
LangChain Blog
酷 壳 – CoolShell
酷 壳 – CoolShell
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
Microsoft Security Blog
Microsoft Security Blog
C
Check Point Blog
MongoDB | Blog
MongoDB | Blog
B
Blog

cs.CL updates on arXiv.org

Playing Along: Learning a Double-Agent Defender for Belief Steering via Theory of Mind RPA-Check: A Multi-Stage Automated Framework for Evaluating Dynamic LLM-based Role-Playing Agents A Triadic Suffix Tokenization Scheme for Numerical Reasoning Hidden Measurement Error in LLM Pipelines Distorts Annotation, Evaluation, and Benchmarking Synthius-Mem: Brain-Inspired Hallucination-Resistant Persona Memory Achieving 94.4% Memory Accuracy and 99.6% Adversarial Robustness on LoCoMo Time is Not a Label: Continuous Phase Rotation for Temporal Knowledge Graphs and Agentic Memory NovBench: Evaluating Large Language Models on Academic Paper Novelty Assessment Policy Split: Incentivizing Dual-Mode Exploration in LLM Reinforcement with Dual-Mode Entropy Regularization METER: Evaluating Multi-Level Contextual Causal Reasoning in Large Language Models Think Before you Write: QA-Guided Reasoning for Character Descriptions in Books METRO: Towards Strategy Induction from Expert Dialogue Transcripts for Non-collaborative Dialogues Retrieval as Generation: A Unified Framework with Self-Triggered Information Planning Do LLMs Know Tool Irrelevance? Demystifying Structural Alignment Bias in Tool Invocations Enhancing Multimodal Large Language Models for Ancient Chinese Character Evolution Analysis via Glyph-Driven Fine-Tuning Polyglot Teachers: Evaluating Language Models for Multilingual Synthetic Data Generation Exploring Knowledge Conflicts for Faithful LLM Reasoning: Benchmark and Method CocoaBench: Evaluating Unified Digital Agents in the Wild MathAgent: Adversarial Evolution of Constraint Graphs for Mathematical Reasoning Data Synthesis Evaluating Memory Capability in Continuous Lifelog Scenario How Robust Are Large Language Models for Clinical Numeracy? An Empirical Study on Numerical Reasoning Abilities in Clinical Contexts Efficient Training for Cross-lingual Speech Language Models Shared Emotion Geometry Across Small Language Models: A Cross-Architecture Study of Representation, Behavior, and Methodological Confounds A Systematic Analysis of the Impact of Persona Steering on LLM Capabilities Uncertainty-Aware Web-Conditioned Scientific Fact-Checking When Valid Signals Fail: Regime Boundaries Between LLM Features and RL Trading Policies When Verification Fails: How Compositionally Infeasible Claims Escape Rejection Mem$^2$Evolve: Towards Self-Evolving Agents via Co-Evolutionary Capability Expansion and Experience Distillation AOP-Smart: A RAG-Enhanced Large Language Model Framework for Adverse Outcome Pathway Analysis OccuBench: Evaluating AI Agents on Real-World Professional Tasks via Language Environment Simulation Advancing Polish Language Modeling through Tokenizer Optimization in the Bielik v3 7B and 11B Series
Towards Understanding the Fragility of Multilingual LLMs ...
Samuele Poppi, Zheng-Xin Yong, Yifei He, Bobbie Chern, Han Zhao, · 2024-10-24 · via cs.CL updates on arXiv.org

Recent advancements in Large Language Models (LLMs) have sparked widespread concerns about their safety. Recent work demonstrates that safety alignment of LLMs can be easily removed by fine-tuning with a few adversarially chosen instruction-following examples, i.e., fine-tuning attacks. We take a further step to understand fine-tuning attacks in multilingual LLMs. We first discover cross-lingual generalization of fine-tuning attacks: using a few adversarially chosen instruction-following examples in one language, multilingual LLMs can also be easily compromised (e.g., multilingual LLMs fail to refuse harmful prompts in other languages). Motivated by this finding, we hypothesize that safety-related information is language-agnostic and propose a new method termed Safety Information Localization (SIL) to identify the safety-related information in the model parameter space. Through SIL, we validate this hypothesis and find that only changing 20% of weight parameters in fine-tuning attacks can break safety alignment across all languages. Furthermore, we provide evidence to the alternative pathways hypothesis for why freezing safety-related parameters does not prevent fine-tuning attacks, and we demonstrate that our attack vector can still jailbreak LLMs adapted to new languages.