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

Blog — PlanetScale
Blog — PlanetScale
Vercel News
Vercel News
奇客Solidot–传递最新科技情报
奇客Solidot–传递最新科技情报
量子位
Y
Y Combinator Blog
IT之家
IT之家
博客园 - 聂微东
L
LangChain Blog
爱范儿
爱范儿
H
Help Net Security
GbyAI
GbyAI
F
Fortinet All Blogs
B
Blog
Microsoft Security Blog
Microsoft Security Blog
罗磊的独立博客
C
Check Point Blog
博客园 - 三生石上(FineUI控件)
小众软件
小众软件
D
DataBreaches.Net
Last Week in AI
Last Week in AI
WordPress大学
WordPress大学
B
Blog RSS Feed
酷 壳 – CoolShell
酷 壳 – CoolShell
宝玉的分享
宝玉的分享

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
An Interpretable N-gram Perplexity Threat Model for Large...
Valentyn Boreiko, Alexander Panfilov, Vaclav Voracek, Matthias H · 2024-10-22 · via cs.CL updates on arXiv.org

A plethora of jailbreaking attacks have been proposed to obtain harmful responses from safety-tuned LLMs. These methods largely succeed in coercing the target output in their original settings, but their attacks vary substantially in fluency and computational effort. In this work, we propose a unified threat model for the principled comparison of these methods. Our threat model checks if a given jailbreak is likely to occur in the distribution of text. For this, we build an N-gram language model on 1T tokens, which, unlike model-based perplexity, allows for an LLM-agnostic, nonparametric, and inherently interpretable evaluation. We adapt popular attacks to this threat model, and, for the first time, benchmark these attacks on equal footing with it. After an extensive comparison, we find attack success rates against safety-tuned modern models to be lower than previously presented and that attacks based on discrete optimization significantly outperform recent LLM-based attacks. Being inherently interpretable, our threat model allows for a comprehensive analysis and comparison of jailbreak attacks. We find that effective attacks exploit and abuse infrequent bigrams, either selecting the ones absent from real-world text or rare ones, e.g., specific to Reddit or code datasets.