About
I am a jointly trained PhD student at Harbin Institute of Technology, Shenzhen, and Peng Cheng Laboratory, admitted in September 2026 and advised by Prof. Jing Li and Prof. Fangming Liu.
I received my BEng degree in Computer Science and Technology from Dalian University of Technology in June 2026. During my undergraduate studies, I was fortunate to be advised by Prof. Liang Zhang, who introduced me to academic research and provided invaluable guidance at the beginning of my research journey. I am sincerely grateful for his mentorship and support.
During my undergraduate studies, I also conducted research at several institutions. At the Shanghai Artificial Intelligence Laboratory, I worked under the guidance of Researcher Jiaxuan Lu.
I later conducted a research internship at the School of Intelligence Science and Technology, Nanjing University, under the supervision of Prof. Ke Xu, with additional guidance from Researcher Xingbo Wang. During this experience, I explored research topics in human-computer interaction. I am sincerely grateful to both of them for their mentorship and support.
I was also fortunate to receive valuable guidance from Prof. Yanbiao Ma at the Gaoling School of Artificial Intelligence, Renmin University of China.
My current research interests focus on internal representation learning and reasoning analysis in multimodal large language models.
News
I mentored a second-year undergraduate intern on open-world KG-Agent exploratory question answering. Our paper was submitted to AAAI 2027 and successfully advanced to Phase 2.
Our paper on agent memory management was accepted to NeurIPS 2026. I am grateful to all collaborators and advisors.
Our paper on personalized alignment for large language models was accepted to EMNLP 2026 Findings. I am grateful to my labmates and advisors.
Our collaborative work on human-computer interaction, to which I contributed as a core team member, was accepted by Visual Informatics.
I led and first-authored two papers submitted to TPAMI; one received a major revision decision, while the other remains under review.
Our paper on modality-decoupled online recursive editing, with me as the first author, was accepted by ICML 2026.
Our paper on semantic-aware knowledge graph completion, with me as the first author, was accepted by Knowledge-Based Systems.
Our paper on temporal knowledge graph forecasting, which I led as the first author, was accepted by ACL 2026.
Our paper on heating load prediction, which I co-first-authored, was accepted by the Journal of Building Engineering.
Our work on multimodal tactile feedback for exercise protection was presented at UbiComp/ISWC 2025.
Selected Publications
View AllMultimodal Knowledge Edit-Scoped Generalization for Online Recursive MLLM Editing
Siyuan Li, Youyuan Zhang, Ruitong Liu, Junxi Wang, Jing Li
Under Review
A scope-aware online editor that controls in-scope cross-modal generalization while preserving out-of-scope locality.
Modality-Decoupled Online Recursive Editing
Siyuan Li, Youyuan Zhang, Fangming Liu, Jing Li
ICML
A modality-decoupled approach to online recursive model editing for MLLMs.
Evolving Beyond Snapshots: Harmonizing Structure and Sequence via Entity State Tuning for Temporal Knowledge Graph Forecasting
Siyuan Li, Yunjia Wu, Yiyong Xiao, Pingyang Huang, Peize Li, Ruitong Liu, Yan Wen, Te Sun
ACL Main Conference
An entity-state tuning framework that harmonizes structural and sequential signals for long-horizon temporal knowledge graph forecasting.
Toward Robust Personalized Alignment for LLMs: Mitigating Persona Drift in Multi-Turn Dialogue
Youyuan Zhang, Siyuan Li, Fangming Liu, Jing Li
EMNLP 2026 Findings
A robust personalized alignment approach for mitigating persona drift in multi-turn dialogue.
Multi-Condition Latent Diffusion Network for Semantic-Aware Knowledge Graph Completion
Siyuan Li, Liang Zhang, Bo Jin, Xiaopeng Wei
Knowledge-Based Systems
A multi-condition latent diffusion network for semantic-aware knowledge graph completion.