A 114k compositional jailbreak dataset is created, generators are fine-tuned for on-the-fly synthesis, and OPTIMUS introduces a continuous evaluator that identifies stealth-optimal regimes missed by binary attack success rates.
Jailjudge: A comprehensive jailbreak judge benchmark with multi-agent enhanced explanation evaluation framework
10 Pith papers cite this work. Polarity classification is still indexing.
10
Pith papers citing it
citation-role summary
background 3
method 1
citation-polarity summary
representative citing papers
PEO optimizes original prompt embeddings continuously over adaptive rounds to jailbreak aligned LLMs, preserving the exact visible prompt text and outperforming discrete suffix, appended embedding, and search-based white-box attacks on harmful-behavior benchmarks.
MTK detects jailbreaks by monitoring the evolution of prompt neighborhood structures on the data manifold through LLM layers, reporting 95% TPR at 5% FPR on benign and 2% on pseudo-malicious prompts plus 85% TPR under adaptive attacks.
A theoretical attacker-defender game in LLM adversarial prompting yields a best-response attack related to existing methods, reveals attacker advantages at equilibrium, and derives a provably optimal defense with stronger empirical performance.
RoTRAG retrieves Rules of Thumb to ground LLM reasoning for harm detection and severity classification in multi-turn dialogues, reporting roughly 40% relative F1 gains and 8.4% lower distributional error on two safety benchmarks while cutting redundant retrieval.
Salami Attack chains low-risk inputs to cumulatively trigger high-risk LLM behaviors, achieving over 90% success on GPT-4o and Gemini while resisting some defenses.
CoT-Guard is a 4B model using SFT and RL that achieves 75% G-mean^2 on hidden objective detection under prompt and code manipulation attacks, outperforming several larger models.
The paper taxonomizes jailbreak attacks and defenses for LLMs, introduces the Security Cube multi-dimensional evaluation framework, benchmarks 13 attacks and 5 defenses, and identifies open challenges in LLM robustness.
Introduces a multi-role red teaming framework using attacker and jury models that increases attack success rates by up to 7.9% on LLM faithfulness in question-answering tasks.
A survey categorizing LLM-powered agent systems into software-based, physical, and hybrid types, covering industrial applications and challenges such as latency and security.
citing papers explorer
-
The Salami Slicing Threat: Exploiting Cumulative Risks in LLM Systems
Salami Attack chains low-risk inputs to cumulatively trigger high-risk LLM behaviors, achieving over 90% success on GPT-4o and Gemini while resisting some defenses.