|
I am a PhD candidate at the School of Computing (SoC), National University of Singapore (NUS) , supervised by Prof. Byran Hooi. My current research focuses on (1) trustworthiness of large language models (llms); (2) llm reasoning (in specialized fields, graphs, tables, etc.); and (3) causal inference. If you are also passionate about these topics, feel free to connect! Email / Github / Google Scholar / Twitter / LinkedIn |
|
|
|
|
|
|
Yuan Sui, Bryan Hooi Preprint, 2026 arXiv This paper proposes a conversation-based framework for self-evolving LLMs through meta-evaluation on non-verifiable learning tasks. |
|
Yuan Sui, Yanming Zhang, Yi Liao, Yu Gu, Guohua Zhang, Zhongqian Sun, Yang Wei, Bryan Hooi ICLR 2026, Workshop AIMS arXiv This paper propose a new thinking paradigm (what-if-analysis) for LLMs to learn to proactive thinking. We test it on Honor of Kings, a globally popular MOBA game. |
|
Tri Cao, Bennett Lim, Yue Liu, Yuan Sui, Yuexin Li, Shumin Deng, Lin Lu, Nay Oo, Shuicheng Yan, Bryan Hooi ICLR, 2026 arXiv This paper introduces VPI-Bench, a benchmark for evaluating visual prompt injection attacks against computer-use agents. |
|
Yuan Sui, Yufei He, Tri Cao, Simeng Han, Yulin Chen, Bryan Hooi ACL Findings, 2026 arXiv / code This paper introduce Meta-Reasoner, a framework that dynamically optimizes inference-time reasoning by enabling LLMs to "think about how to think". |
|
Yulin Chen, Haoran Li, Yuan Sui, Yue Liu, Yufei He, Yangqiu Song, Bryan Hooi ACL Findings, 2026 arXiv This paper proposes a defense against prompt injection attacks by referencing the executed instruction to improve robustness of LLMs. |
|
Yi Liao, Yu Gu, Yuan Sui, Guohua Tang, Zhongqian Sun, Wei Yang Technical Report @ Tencent, 2025 arXiv This paper explores how we could leverage LLMs in dynamic, complex game environment. We use RL techniques to train the LLM to interact with game environment to acquire deep understanding of the underlying game mechanisms. |
|
Yuan Sui, Yufei He, Zifeng Ding, Bryan Hooi ACL, 2025 arXiv / code This paper introduces OKGQA, a benchmark for evaluating Knowledge Graph-enhanced LLMs in open-ended question answering, focusing on reducing hallucinations and improving reasoning. |
|
Yuan Sui, Yufei He, Nian Liu, Xiaoxin He, Kun Wang, Bryan Hooi ACL, 2025 arXiv / code This paper presents FiDeLiS, a method combining knowledge graphs and LLMs to enhance reasoning efficiency and reduce hallucinations in downstream QA tasks. |
|
Yulin Chen*, Haoran Li*, Yuan Sui, Yufei He, Yue Liu, Yangqiu Song, Bryan Hooi ACL, 2025 arXiv This paper investigates detecting and removing indirect prompt injection attacks on LLMs. |
|
Yulin Chen, Haoran Li, Yuan Sui, Yangqiu Song, Bryan Hooi EMNLP, 2025 arXiv This paper demonstrates that backdoor-powered prompt injection attacks can nullify existing defense methods for LLMs. |
|
Yufei He, Ruoyu Li, Alex Chen, Yue Liu, Yulin Chen, Yuan Sui, Cheng Chen, Yi Zhu, Luca Luo, Frank Yang, Bryan Hooi EMNLP, 2025 arXiv This paper enables self-improving agents to learn at test time with human-in-the-loop guidance. |
|
Yufei He, Yuan Sui, Xiaoxin He, Yue Liu, Yifei Sun, Bryan Hooi WWW, 2025 (oral) arXiv / code This paper introduces UniGraph2, a foundation model that unifies multimodal information and graph structures for effective representation learning on multimodal graphs. |
|
Yufei He, Yuan Sui, Xiaoxin He, Bryan Hooi KDD, 2025 arXiv / code This paper introduces UniGraph, a framework for cross-domain graph learning using Text-Attributed Graphs (TAGs) and a proposed pre-training algorithm for effective zero-shot and few-shot learning. |
|
Yuan Sui*, Jiaru Zou*, Mengyu Zhou, Xinyi He, Lun Du, Shi Han, Dongmei Zhang EMNLP, 2024 arXiv / code / poster / slide TAP4LLM presents a versatile pre-processor suite for leveraging LLMs in table-based tasks effectively. |
|
Yuan Sui, Mengyu Zhou, Mingjie Zhou, Shi Han and Dongmei Zhang WSDM, 2024 arXiv / code / blog / poster / slide This paper releases SUC benchmark along with comprehensive empirical studies to evaluate and detect the structural understanding capabilities of LLMs on table data. |
|
Wei Yan, Yu Shi, Zengyan Ji, Yuan Sui, Zhenzhen Tian, Wanjing Wang, Qiushi Cao Engineering Applications of Artificial Intelligence, 2023 (IF=8) This paper proposes a virtual knowledge graph for the digital modeling and intelligent predicate maintenance of hydraulic systems in manufacturing. |
|
Yuan Sui, Shanshan Feng, Huaxiang Zhang, Jian Cao, Liang Hu, Nengjun Zhu Knowledge-Based Systems (KBS), 2022 (IF=8.139) This paper propose CF-KGQA to improve KGQA tasks by address spurious relations through causal interference. |
|
Yuan Sui, Fanyang Bu, Yingting Hu, Wei Yan, Liang Zhang IJCNN, 2022 (oral) Trigger-GNN proposes entity triggers to improve the nested NER task. |
|
|
:
:
:
:
|
|
|
|
|
Thanks Jon Barron
for this amazing template.
|