Qiusi Zhan | 詹秋思
I am a Ph.D. student at the Univerity of Illinois Urbana Champaign (UIUC), advised by Prof. Daniel Kang. Previously, I obtained my master's degree from UIUC, advised by Prof. Heng Ji, and completed my bachelor's degree from Peking University, advised by Prof. Sujian Li.
Email /
CV /
Google Scholar /
Github
|
|
Research
I have a broad research interest in natural language processing, currently
centered on enhancing the robustness and trustworthiness of (multimodal) large language models (LLMs) and LLM agents. I also have experience in the fields of event extraction and dialog systems.
|
|
Adaptive Attacks Break Defenses Against Indirect Prompt Injection Attacks on LLM Agents
Qiusi Zhan, Richard Fang, Henil Shalin Panchal, Daniel Kang
NAACL Findings, 2025
We test eight different defenses of Indirect Prompt Injection attacks and demonstrate that they are vulnerable to adaptive attacks.
|
|
MM-PoisonRAG: Disrupting Multimodal RAG with Local and Global Poisoning Attacks
Hyeonjeong Ha*, Qiusi Zhan*, Jeonghwan Kim , Dimitrios Bralios, Saikrishna Sanniboina, Nanyun Peng, Kai-wei Chang, Daniel Kang, Heng Ji
arXiv, 2025
We propose MM-PoisonRAG, a novel knowledge poisoning attack framework for multimodal RAG.
|
|
INJECAGENT: Benchmarking Indirect Prompt Injections in
Tool-Integrated Large Language Model Agents
Qiusi Zhan, Zhixiang Liang, Zifan Ying, Daniel Kang
ACL Findings, 2024
We benchmark IPI attacks in tool-integrated LLM agents and show that most of the agents are vulnerable to such attacks.
|
|
Removing RLHF Protections in GPT-4 via Fine-Tuning
Qiusi Zhan, Richard Fang, Rohan Bindu, Akul Gupta, Tatsunori Hashimoto, Daniel Kang
NAACL, 2024
We demonstrate that fine-tuning GPT-4 with just 340 examples can subvert its RLHF protections, underscoring the critical need for enhanced security measures.
|
|
GLEN: General-Purpose Event Detection for Thousands of Types
Qiusi Zhan*, Sha Li*,
Kathryn Conger
, Martha Palmer
, Heng Ji
, Jiawei Han
EMNLP, 2023
We build a wide‑coverage event detection dataset, with 205K event mentions covering 3,465 types and develop a novel multi‑stage event detection model.
|
|
User Simulator Assisted Open-ended Conversational Recommendation System
Qiusi Zhan,
Xiaojie Guo,
Heng Ji,
Lingfei Wu
ACL Workshop, 2023
We fine-tune pre-trained CRS based on its interaction with a designed User Simulator using reinforcement learning to largely improve its recommendation performance.
|
|
EA2E: Improving Consistency with Event Awareness for Document-Level Argument Extraction
Qi Zeng*,
Qiusi Zhan*,
Heng Ji
NAACL Findings, 2022
We improve the argument consistency of multiple events in documents by making the model aware of event relations.
|
|
ConFiguRe: Exploring Discourse-level Chinese Figures of Speech
Dawei Zhu,
Qiusi Zhan,
Zhejian Zhou,
Yifan Song,
Jiebin Zhang,
Sujian Li
COLING, 2022
We benchmark the detection of Chinese Figures of Speech.
|
 |
University of Illinois Urbana-Champaign, IL
2023.08 - Present
Ph.D. Student in Computer Science
Advisor: Prof. Daniel Kang
|
 |
University of Illinois Urbana-Champaign, IL
2021.08 - 2022.12
Master in Electronical and Computer Engineering
Advisor: Prof. Heng Ji
|
 |
Peking University, China
2017.09 - 2021.07
B.S. in Computer Science
Advisor: Prof. Sujian Li
|
 |
University of California Santa Barbara, CA
2020.07 - 2020.09
Visiting Research Student
Advisor: Prof. Xifeng Yan
|
 |
Microsoft, WA
2024.05 - 2024.08
Applied Scientist Intern
Mentor: Yu Hu
|
 |
JD.COM Silicon Valley Research Center, CA
2022.01 - 2022.05
Research Scientist Intern
Mentor: Lingfei Wu
|
 |
ByteDance, China
2021.04 - 2021.07
Natural Language Processing Engineer Intern
Mentor: Bo Zhao
|
This website is based on a template created by Jon Barron.
Last updated: Mar. 12, 2025
|
|