The integration of additional modalities increases the susceptibility of large vision-language models (LVLMs) to safety risks, such as jailbreak attacks, compared to their language-only counterparts. While existing research primarily focuses on post-hoc alignment techniques, the underlying safety mechanisms within LVLMs remain largely unexplored. In this work , we investigate whether LVLMs inherently encode safety-relevant signals within their internal activations during inference. Our findings reveal that LVLMs exhibit distinct activation patterns when processing unsafe prompts, which can be leveraged to detect and mitigate adversarial inputs without requiring extensive fine-tuning. Building on this insight, we introduce HiddenDetect, a novel tuning-free framework that harnesses internal model activations to enhance safety. Experimental results show that HiddenDetect surpasses state-of-the-art methods in detecting jailbreak attacks against LVLMs. By utilizing intrinsic safety-aware patterns, our method provides an efficient and scalable solution for strengthening LVLM robustness against multimodal threats. Our code and data will be released publicly.

Non-classical computation slicing is an abstraction mechanism for pruning state space and it plays an important role in tackling state explosion problem in predicate detection. This paper shows that the slicing technique based on prime event structures can improve the expressiveness of that abstract method and extend its application scope greatly. We introduce the slicing idea into the area of event structure and develop a complete theory framework with related techniques for event structure slicing. This slicing technique can serve as a useful supplement and high level extension to the existent computation slicing which can not handle the situations with conflicts or execution choices.

Kleene algebra with tests (KAT) is a prominent specification language used for formalizing the behavior of non-deterministic structured programs generally called regular programs. Regular programs with probabilistic information have richer and more powerful expressiveness than normal regular programs. However, KAT is incapable of specifying such recently widely used programs. We construct a complete theory of probabilistic Kleene algebra with tests (PKAT) for reasoning about regular programs with probability. We offer a model termed probabilistic configuration transition systems, whose states are configurations composed of a pair of a PKAT expression and a data-state. To determine the transition relation in the model, we define an operational semantics for PKAT, and establish a probabilistic bisimulation equivalence relation in PKAT. The soundness of the equations of PKAT is validated with respect to the bisimulation, to find equivalences without calculating the actual bisimulation relation.