About Me

I am a Ph.D. candidate in Computer Science and Engineering at The Chinese University of Hong Kong (CUHK), supervised by Prof. Irwin King.

Previously, I received an M.S. in Computer Science and Technology from Harbin Institute of Technology under the supervision of Prof. Philippe Fournier-Viger, and visited the National University of Singapore in 2025, working closely with Prof. Tat-Seng Chua. I also have a close collaboration with Prof. Rex Ying at Yale University.

My research focuses on personalized and structure-aware foundation models, with an emphasis on non-Euclidean geometry, user preferences, and long-term memory.

Research Interests

  • Non-Euclidean geometry: hyperbolic learning, structural priors, and multimodal continual learning.
  • Personalized foundation models: preference modeling, representation-space personalization, and efficient adaptation.
  • Agents, memory, and retrieval: long-term personalization and user-specific knowledge.

I welcome conversations and collaborations on these topics. Feel free to contact me!

News 🔥

Publications

Selected papers are listed below. See Google Scholar for the full list. * denotes equal contribution; † denotes corresponding author.

A Survey on Personalized Large Language Models

Jiahong Liu, Zexuan Qiu, Zhongyang Li, Quanyu Dai, Wenhao Yu, Jieming Zhu, Minda Hu, Menglin Yang, Tat-Seng Chua, Irwin King

ACM Computing Surveys (IF: 30.4)   Paper ↗   ·   Project ↗   ·   Resources ↗

Personalized prompting, adaptation, alignment, evaluation, and future directions.

ACM CSUROne-size-fits-all responses compared with personalized LLM responses

Memory Has Geometry: Non-Uniform Geometric Memory for Long-Horizon Personalized AI

Jiahong Liu, Wenhao Yu, Zexuan Qiu, Menglin Yang, Irwin King

ICDM 2026 · Blue Sky Ideas Track (Oral)   Paper ↗

A geometric perspective on long-term personalized memory, modeled as a user-specific state space with locally heterogeneous geometry.

ICDM 2026Conceptual framework from Memory Has Geometry

FlatLand: Personalized Graph Federated Learning via Tailored Lorentz Space

Jiahong Liu, Ram Samarth B B, Xinyu Fu, Menglin Yang, Weixi Zhang, Rex Ying, Irwin King

ICML 2026 (Oral, 0.7%)   Paper ↗   ·   arXiv ↗   ·   Code ↗

Tailored Lorentz spaces separate client-specific heterogeneity from shared knowledge for personalized graph federated learning.

ICML 2026 · Oral (0.7%)FlatLand framework for personalized federated learning in tailored Lorentz spaces

Hyperbolic Multimodal Continual Learning

Jiahong Liu, Ming Shen, Xiaohao Liu, Rex Ying, Menglin Yang, Tat-Seng Chua, Irwin King

ICML 2026   Paper ↗   ·   Code ↗

Preserving semantic relations and hierarchical geometry while adapting multimodal models to a sequence of new tasks.

ICML 2026HMCL preserves semantic similarity and hierarchy under a shared hyperbolic isometry

PerFit: Exploring Personalization Shifts in Representation Space of LLMs

Jiahong Liu, Wenhao Yu, Quanyu Dai, Zhongyang Li, Jieming Zhu, Menglin Yang, Tat-Seng Chua, Irwin King

ICLR 2026   Paper ↗   ·   Code ↗ Also at KnowFM @ ACL 2025

Shared and user-specific representation shifts enable effective LLM personalization with 92.3% fewer parameters than prior state-of-the-art methods.

ICLR 2026PerFit learns shared and user-specific shifts in the representation space of LLMs

Position: Beyond Euclidean — Foundation Models Should Embrace Non-Euclidean Geometries

Neil He*, Jiahong Liu*, Buze Zhang, Ngoc Bui, Ali Maatouk, Menglin Yang, Irwin King, Melanie Weber, Rex Ying

LoG 2025 (Oral)   Paper ↗ * Equal contribution

A roadmap for incorporating the geometry of real-world data into the design of foundation models.

LoG 2025Hyperbolic, Euclidean, and spherical geometries model different token relationships

Hyperbolic Fine-Tuning for Large Language Models

Menglin Yang, Ram Samarth B B, Aosong Feng, Bo Xiong, Jiahong Liu, Irwin King, Rex Ying

NeurIPS 2025 (Spotlight)   Paper ↗   ·   Project ↗   ·   Code ↗

HypLoRA uses hyperbolic low-rank adaptation to exploit latent hierarchical structure in large language models.

NeurIPS 2025HypLoRA illustration of token frequency, embedding norms, and hierarchical geometry

StepKV: Step-Aware KV Cache Compression for LLM Agents

Boyu Feng, Jiahong Liu†, Yifan Li, Wenhao Yu, Zexuan Qiu, Yuliang Sun, Ming Shen, Xiang Li, Quanyu Dai, Irwin King

Preprint 2026   Paper ↗   † Corresponding author

Hyperbolic Multimodal Continual Learning: A Closest-Admissible Solution

Jiahong Liu, Ming Shen, Xiaohao Liu, Rex Ying, Menglin Yang, Tat-Seng Chua, Irwin King

Preprint 2026   Paper ↗

Discovering Representative Attribute-stars via Minimum Description Length

Jiahong Liu, Min Zhou, Philippe Fournier-Viger, Menglin Yang, Lujia Pan, Mourad Nouioua

ICDE 2022 (Oral)   Paper ↗

CSPM: Discovering Compressing Stars in Attributed Graphs

Jiahong Liu, Philippe Fournier-Viger, Min Zhou, Ganghuan He

Information Sciences   Paper ↗

Enhancing Hyperbolic Graph Embeddings via Contrastive Learning

Jiahong Liu, Menglin Yang, Min Zhou, Shanshan Feng, Philippe Fournier-Viger

SSL @ NeurIPS 2021 (Oral)

Honors and Awards

  • PhD IMPAC Award, The Chinese University of Hong Kong.
  • Excellent Master’s Thesis Award (top 5%), Harbin Institute of Technology.

Academic Service

Workshops

Tutorials

Reviewing

  • Conferences: ICML (2025–), ICLR (2025–), NeurIPS (2024–), KDD (2024–), WebConf (2026–), AISTATS (2025–), and IJCNN (2023–).
  • Journals: Artificial Intelligence, IEEE Transactions on Artificial Intelligence, Neural Networks, Pattern Recognition, Information Sciences, Expert Systems with Applications, Neurocomputing, Decision Support Systems, and AIP Advances.

Invited Talks

  • 2025, King’s College London, “Personalized Large Language Models.”

Teaching

Teaching assistant for the following courses:

  • The Chinese University of Hong Kong: Formal Languages and Automata Theory (Fall 2022); Data Structures (Spring 2023).
  • Harbin Institute of Technology: Data Mining (Fall 2020); Information Retrieval (Spring 2020).

Last updated: November 2026. Built with AcadHomepage.