Helpful to a Fault: Measuring Illicit Assistance in Multi-Turn, Multilingual LLM Agents
cs.CL, cs.LG
Submitted: 2026-02-18
Updated: 2026-11-03
Code: https://github.com/epfl-nlp/helpful-to-a-fault
Terminology
Sources
- Breach By A Thousand Leaks: Unsafe Information Leakage in `Safe' AI Responses
- The Llama 3 Herd of Models
- MAPS: A Multilingual Benchmark for Agent Performance and Security
- OS-Harm: A Benchmark for Measuring Safety of Computer Use Agents
- DrAttack: Prompt Decomposition and Reconstruction Makes Powerful LLM Jailbreakers
- DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
- Jailbreaking ChatGPT via Prompt Engineering: An Empirical Study
- Multilingual E5 Text Embeddings: A Technical Report
- Low-Resource Languages Jailbreak GPT-4
- Monitoring Decomposition Attacks in LLMs with Lightweight Sequential Monitors
- Siren: A Learning-Based Multi-Turn Attack Framework for Simulating Real-World Human Jailbreak Behaviors
Related papers
- ALARM: Audio-Language Alignment for Reasoning Models
- ThinkSum: Probabilistic reasoning over sets using large language models
- Coherence boosting: When your pretrained language model is not paying enough attention
- From Facts to Insights: A Persona-Driven Dual Memory Framework and Dataset for Role-Playing Agents
- Improving General Role-Playing Agents via Psychology-Grounded Reasoning and Role-Aware Policy Optimization
- The Art of Asking: Multilingual Prompt Optimization for Synthetic Data