214 citations · 1.5k across the 43 of their papers we have counts for
12 papers · 1 filter
Where Should a Stuck Agent Spend Its Next Tokens? Causal Attribution of Inference-Time Resources
Saiyue Lyu, Jerrold Huang, Yunbei Zhang +4
Inference-time scaling is often treated as a token-budget problem, but a failed coding agent can seek qualitatively different resources: computation, information, or feedback. We i…
SkillScreen: High-Recall Screening of Malicious Agent Skills with Decision Models
Jerrold Huang, Junwei Quan, Saiyue Lyu +2
Agent skills turn natural-language instructions into actions, creating a security challenge: legitimate automation and malicious behaviour can look deceptively similar. A practical…
VERSE: Verified Self-Evolving Optimizer for Agent Harnesses
Zekai Wang, Yingqiang Ge, Zekun Wang +6
Harness evolution improves an LLM agent's prompts, tools, and workflow, while the optimizer's own tools and procedures often remain fixed. We study whether an optimizer can improve…
Safety Must Survive Self-Improvement: Why Failures Persist and How Agents Recover
Yunbei Zhang, Janet Wang, Saiyue Lyu +5
Recursive self-improvement (RSI) allows agents to carry useful changes across generations. Maintaining safety across these generations involves both preventing unsafe behavior from…
Deny Without Disabling: Authorization-Paired Evaluation and Control for Multi-Agent Systems
Yunbei Zhang, Saiyue Lyu, Janet Wang +4
Multi-agent systems derive their capabilities from sharing evidence, delegating tasks, and combining information across agents. The same process creates a safety problem: contribut…
DRIFTLENS: Measuring Memory-Induced Reasoning Drift in Personalized Language Models
Xi Fang, Weijie Xu, Yingqiang Ge +3
Personalization changes what a model says to a user; we show that it can also change the reasoning trajectory used to justify the response. Modern LLMs personalize interactions by…
One Skill, Two Masters: Serving Users and Attackers through Skill Injection
Haochang Hao, Dehai Min, Zhifang Zhang +5
Large language model agents use skills that package reusable instructions, code, and resources. Existing skill-injection attacks exploit trust in these packages by adding malicious…
Stop Comparing LLM Agents Without Disclosing the Harness
Yunbei Zhang, Janet Wang, Yingqiang Ge +3
This position paper argues that, for long-horizon tasks evaluated across models with comparable frontier capability, the agent execution harness, namely the infrastructure layer th…
ClawSafety: "Safe" LLMs, Unsafe Agents
Bowen Wei, Yunbei Zhang, Jinhao Pan +5
Personal AI agents like OpenClaw run with elevated privileges on users' local machines, where a single successful prompt injection can leak credentials, redirect financial transact…
SafeCRS: Personalized Safety Alignment for LLM-Based Conversational Recommender Systems
Haochang Hao, Yifan Xu, Xinzhuo Li +2
Current LLM-based conversational recommender systems (CRS) primarily optimize recommendation accuracy and user satisfaction. We identify an underexplored vulnerability in which rec…
Agents in the Wild: Safety, Society, and the Illusion of Sociality on Moltbook
Yunbei Zhang, Kai Mei, Ming Liu +5
We present the first large-scale empirical study of Moltbook, an AI-only social platform where 27,269 agents produced 137,485 posts and 345,580 comments over 9 days. We report thre…
Visual Exclusivity Attacks: Automatic Multimodal Red Teaming via Agentic Planning
Yunbei Zhang, Yingqiang Ge, Weijie Xu +3
Current multimodal red teaming treats images as wrappers for malicious payloads via typography or adversarial noise. These attacks are structurally brittle, as standard defenses ne…