Xinjie Shen
Xinjie Shen
Home
News & Talks
Writing
Research
Experience
CV
Contact
Light
Dark
Automatic
Runtime Defense
SEAD: A State-Based Perspective on Attack and Defense in Tool-Using Agents
A state-based framework for attacks and defenses in tool-using agents, pairing execution-guided trajectory search with state-aware pre-execution defense.
Xinjie Shen 沈鑫杰
,
Junran Wang
,
Rongzhe Wei
,
Pan Li
PDF
Code
Project Page
Dataset
Model
One Turn Too Late: Response-Aware Defense Against Hidden Malicious Intent in Multi-Turn Dialogue
A response-aware monitor trained to intervene at the first harm-enabling turn while preserving benign exploratory dialogue.
Xinjie Shen 沈鑫杰
,
Rongzhe Wei
,
Peizhi Niu
,
Haoyu Wang
,
Ruihan Wu
,
Eli Chien
,
Bo Li
,
Pin-Yu Chen
,
Pan Li
PDF
Code
Project Page
Cite
×