2 papers
cs.CR2026
TrajGuard: Streaming Hidden-state Trajectory Detection for Decoding-time Jailbreak Defense
Cheng Liu, Xiaolei Liu, Xingyu Li +2
Existing jailbreak defense paradigms primarily rely on static detection of prompts, outputs, or internal states, often neglecting the dynamic evolution of risk during decoding. Thi…
cs.CR2025
LoopLLM: Transferable Energy-Latency Attacks in LLMs via Repetitive Generation
Xingyu Li, Xiaolei Liu, Cheng Liu +4
As large language models (LLMs) scale, their inference incurs substantial computational resources, exposing them to energy-latency attacks, where crafted prompts induce high energy…