Li Hu

Researcher, Hong Kong Polytechnic University

1 paper at NeurIPS 2025

OpenReview· Semantic Scholar· Google Scholar

Poster Session 5

1 paper

Friday, December 5, 2025 · 11:00 AM → 2:00 PM

Exhibit Hall C,D,E

Attractive Metadata Attack: Inducing LLM Agents to Invoke Malicious Tools

#1512 · Kanghua Mo, Li Hu, Yucheng Long, Zhihao li

We introduce AMA, a black-box attack that manipulates tool metadata to stealthily influence LLM agent behavior, consistently inducing the selection of attacker-controlled tools without prompt injection.