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.
Learningsafetyconstraintsforlarge language models
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
citation-role summary
background 1
citation-polarity summary
years
2026 4roles
background 1polarities
unclear 1representative citing papers
Sparse autoencoders isolate unstable features in reward model representations and enable two mitigation techniques that reduce preference errors on perturbed inputs without retraining.
GCD uses diffusion model priors to guide suffix search, achieving higher attack success rates with better semantic adherence and lower detection than GCG-style methods.
The paper proposes a bidirectional continuum between LLMs and control systems, covering LLM-assisted controller design, control-based LLM steering, and state-space modeling of LLMs.
citing papers explorer
-
Defending Jailbreak Attacks on Large Language Models via Manifold Trajectory Kinetics
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.
-
Preference Instability in Reward Models: Detection and Mitigation via Sparse Autoencoders
Sparse autoencoders isolate unstable features in reward model representations and enable two mitigation techniques that reduce preference errors on perturbed inputs without retraining.
-
Greedy Coordinate Diffusion: Effective and Semantically Coherent Adversarial Attacks via Diffusion Guidance
GCD uses diffusion model priors to guide suffix search, achieving higher attack success rates with better semantic adherence and lower detection than GCG-style methods.
-
When control meets large language models: From words to dynamics
The paper proposes a bidirectional continuum between LLMs and control systems, covering LLM-assisted controller design, control-based LLM steering, and state-space modeling of LLMs.