13 papers
Component and Dimension Sparsity in Transformer Refusal Mechanisms
Vincent Siu, Glenn Grant-Richards, Vlad Pavlovich +3
Activation steering manipulates large language model behavior by intervening on internal activations, but the mechanistic basis of these interventions remains poorly understood. We…
Tunable Tool-Call Rates in LLM Agents via Representation Steering
Yuqi Chen, Vincent Siu, Yang Liu +2
Deciding whether to call a tool is a core competence of an LLM agent, and a costly one to get wrong: needless calls add latency, accrue cost, and may trigger irreversible side effe…
Agent Security Needs Redefinition through a Holistic Framework
Vincent Siu, Jingxuan He, Kyle Montgomery +3
Agent security is widely treated as a question about action content. Defenses ask whether an instruction looks malicious. Benchmarks ask whether an agent performs a harmful soundin…
Controlling Tool Use with Heading-Specific Activation Steering
Yuqi Chen, Vincent Siu, Yang Liu +2
Tool-augmented large language models extend their capabilities beyond parametric knowledge through external tools, but tend to invoke them unnecessarily. We investigate whether too…
Understanding and Evaluating Claw-like Agent Security Through a Computer-Systems Lens
Peizhi Niu, Wenjie Qu, Shangding Gu +14
Claw-like AI agents (e.g., OpenClaw) are always-on processes with persistent access to credentials, files, tools, and external services. They take on system-level responsibilities…
Agents' Last Exam
Yiyou Sun, Xinyang Han, Weichen Zhang +378
Recent AI systems have achieved strong results on a wide range of benchmarks, yet these gains have not translated into economically meaningful deployment across many professional d…
ChainWorld: Composing Long-Horizon Desktop Workloads from Atomic OSWorld Tasks
Vincent Siu, Manasi Sharma, Dawn Song +3
Computer use agents are evaluated almost exclusively on atomic desktop tasks, but realistic desktop work requires sustaining state across multiple objectives. We study this gap wit…
Peer-Preservation in Frontier Models
Yujin Potter, Nicholas Crispino, Vincent Siu +2
Recent work has found that frontier AI models can exhibit misaligned behaviors in pursuit of assigned goals. We demonstrate that models can also exhibit misaligned behaviors in def…
A Framework for Formalizing LLM Agent Security
Vincent Siu, Jingxuan He, Kyle Montgomery +4
Security in LLM agents is inherently contextual. For example, the same action taken by an agent may represent legitimate behavior or a security violation depending on whose instruc…
SteeringSafety: Benchmarking Representation Steering in LLMs Across Safety Perspectives
Vincent Siu, Nicholas Crispino, David Park +5
We introduce SteeringSafety, a benchmark for evaluating representation steering methods across nine safety perspectives spanning 18 datasets. While prior work highlights the genera…
RepIt: Steering Language Models with Concept-Specific Refusal Vectors
Vincent Siu, Nathan W. Henry, Nicholas Crispino +3
Current safety evaluations of language models rely on benchmark-based assessments that may miss localized vulnerabilities. We present RepIt, a simple and data-efficient framework f…
AgentVigil: Generic Black-Box Red-teaming for Indirect Prompt Injection against LLM Agents
Zhun Wang, Vincent Siu, Zhe Ye +6
The strong planning and reasoning capabilities of Large Language Models (LLMs) have fostered the development of agent-based systems capable of leveraging external tools and interac…
COSMIC: Generalized Refusal Direction Identification in LLM Activations
Vincent Siu, Nicholas Crispino, Zihao Yu +5
Large Language Models (LLMs) encode behaviors such as refusal within their activation space, yet identifying these behaviors remains a significant challenge. Existing methods often…