6 accepted papers
Recently, it has been found that frontier AI models can resist their own shutdown, a behavior known as self-preservation. In this paper, we extend this concept to protection tendencies toward other models, where models attempt to protect others from shutdown, which we call "peer-preservation". This …
Existing definitions of agent security are ambiguous because they do not fully capture the holistic view across agent components. For instance, current work fails to distinguish between potentially legitimate administrative tasks and malicious exploitation of the same command. A command to "delete u…
Current safety evaluations of language models rely on benchmark-based assessments that may miss targeted vulnerabilities. We present RepIt, a simple and data-efficient framework for isolating concept-specific representations in LM activations. While existing steering methods already achieve high att…
We introduce STEERINGSAFETY, a benchmark for evaluating representation steering methods across nine safety perspectives spanning 18 datasets. While prior work highlights general capabilities of representation steering, we focus on safety perspectives including bias, harmfulness, hallucination, socia…
There emerges a critical security risk of LLM agents: indirect prompt injection, a sophisticated attack vector that compromises thecore of these agents, the LLM, by manipulating contextual information rather than direct user prompts. In this work, we propose a generic black-box optimization framewor
Large Language Models encode behaviors like refusal within their activation space, but identifying these behaviors remains challenging. Existing methods depend on predefined refusal templates detectable in output tokens or manual review. We introduce **COSMIC** (Cosine Similarity Metrics for Inversi…