AI & ML interests
Integrated sensing and communication, Physical AI for human motion and health, security and privacy of wireless, sensing and AI systems, trustworthy AI and LLM safety.
Recent Activity
Papers
How Narrative Wrapping Affects LLM Refusal: A Cross-Language Benchmark and Defense
MUGEN: A Unified Framework for Efficient Motion Understanding and Generation
XL-Stack Lab
One stack, from integrated sensing and communication up to foundation models for Physical AI.
XL-Stack Lab is a research group in the Department of Computer Science at Florida State University, led by Prof. Xin Liu. We build full-stack systems that run from integrated sensing and communication and NextG wireless networks, through edge intelligence, up to foundation models for Physical AI.
Research areas
- Integrated sensing and communication (Wi-Fi, mmWave, NextG systems)
- Physical AI for human motion and health
- Security and privacy of wireless, sensing and AI systems, including the safety of language models
Datasets
GUISE (gated)
A cross-language benchmark of matched bare and narrative-wrapped harmful requests, in English, modern Chinese and Classical Chinese, from the paper How Narrative Wrapping Affects LLM Refusal: A Cross-Language Benchmark and Defense.
- Dataset: xl-stack-lab/GUISE
- Project page: https://jye16.github.io/guise/
- Code (AXIS defense and evaluation): https://github.com/JYe16/GUISE_Official
The dataset contains harmful requests and attack prompts, so access is gated. Requests are reviewed by hand and granted to individual researchers for language-model safety research. Fill in the form on the dataset page; reviews usually take a few days.
Citation
If you use our datasets, please cite the paper listed on the dataset page.