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

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.

We test eight different defenses of Indirect Prompt Injection attacks and demonstrate that they are vulnerable to adaptive attacks.

We propose MM-PoisonRAG, a novel knowledge poisoning attack framework for multimodal RAG.

We benchmark IPI attacks in tool-integrated LLM agents and show that most of the agents are vulnerable to such attacks.

We demonstrate that fine-tuning GPT-4 with just 340 examples can subvert its RLHF protections, underscoring the critical need for enhanced security measures.

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.

We fine-tune pre-trained CRS based on its interaction with a designed User Simulator using reinforcement learning to largely improve its recommendation performance.

We improve the argument consistency of multiple events in documents by making the model aware of event relations.

We benchmark the detection of Chinese Figures of Speech.

This website is based on a template created by Jon Barron.
Last updated: Mar. 12, 2025