Zhishu LiuAcademic homepage

Hello! I'm

Zhishu Liu

Small expressions.
Interesting questions.

I study micro-expressions and facial action units, with an interest in multimodal large models for affective computing.

Explore my research Google Scholar
Micro-expressionsAction unitsMultimodal models
My Pikachu avatar in a sunny forest
Portrait of Zhishu Liu
MSc Student · Computer ScienceCity University of Hong Kong (Dongguan) · Visiting student at Great Bay University
Bachelor’s degree · Artificial IntelligenceCollege of Artificial Intelligence · Jilin University2021.09 – 2025.06 · Advised by Prof. Tieru Wu

A little about me

Research & curiosity

I am Zhishu Liu, currently an MSc student in Computer Science at City University of Hong Kong (Dongguan). I received my bachelor's degree from the College of Artificial Intelligence, Jilin University, advised by Prof. Tieru Wu, and I am also a visiting student at Great Bay University under the supervision of Prof. Zitong Yu.

My research focuses on micro-expression recognition, micro-AU recognition, and multimodal large models for affective computing. I hope to pursue research that is both interesting and meaningful.

I have also served as a reviewer for Pattern Recognition.

2025.06 · News

Our paper “AULLM” was accepted by CCBR 2025 (The 19th Chinese Conference on Biometric Recognition), with an oral presentation.

First-author work

First-author publications

Overview of the AULLM++ micro-expression action unit detection frameworkAULLM++
TBIOM · Under reviewMinor Revision2026

AULLM++: Structured-Token-Conditioned Large Language Models for Micro-Expression Action Unit Detection

Zhishu Liu, Kaishen Yuan, Bo Zhao, Hui Ma, Zitong Yu.

Under review at IEEE Transactions on Biometrics, Behavior, and Identity Science (TBIOM); Minor Revision received on October 5, 2026 (not yet accepted).

Structured-token-conditioned large language models for micro-expression action unit detection.

Google Scholar Project
Overall SAGE-IMER framework with anatomical quotient pooling, emotion-fiber graph filtering, and strength–activation decouplingSAGE-IMER
International Journal of Computer Vision (IJCV)Under Review2026

Beyond Forgetting: From Local Evidence to Global Competition in Incremental Micro-Expression Recognition

Zhishu Liu, Hui Ma, Kaishen Yuan, Bo Zhao, Zeheng Wang, Zitong Yu, Xiaopeng Hong.
Zhishu Liu, Hui Ma, and Kaishen Yuan contributed equally.

Submitted to IJCV in September 2026; currently under review.

Anatomical motion pooling, emotion-aware classifier graph filtering, and decoupled strength–activation control for incremental micro-expression recognition.

Overall GSCL framework with canonical anatomical alignment, cross-source AU prototype pulling, and spectral graph alignmentGSCL
IEEE Transactions on Multimedia (TMM)Under Review2026

Geometric and Spectral Consistency Learning for Cross-Dataset Micro-Expression Action Unit Detection

Zhishu Liu, Kaishen Yuan, Bo Zhao, Hui Ma, Dan Guo, Zitong Yu.
Zhishu Liu and Kaishen Yuan contributed equally.

Submitted to TMM in June 2026; currently under review.

Canonical anatomical alignment and AU prototype spectral alignment for cross-dataset micro-expression action unit detection.

Overview of the AU-LLM enhanced feature fusion frameworkAU-LLM
CCBR 2025Oral presentation2025

AU-LLM: Micro-Expression Action Unit Detection via Enhanced LLM-Based Feature Fusion

Zhishu Liu, Kaishen Yuan, Bo Zhao, Yong Xu, Zitong Yu.

Enhanced LLM-based feature fusion for micro-expression action unit detection.

Google Scholar Project

Non-first-author work

Collaborative publications

Overall framework of AffectAgent with collaborative planning, retrieval, evidence filtering, and emotion generationAffectAgent
ACM Multimedia 2026Accepted2026

AffectAgent: Collaborative Multi-Agent Reasoning for Retrieval-Augmented Multimodal Emotion Recognition

Zeheng Wang, Zitong Yu, Yijie Zhu, Bo Zhao, Haochen Liang, Taorui Wang, Wei Xia, Jiayu Zhang, Zhishu Liu, Hui Ma, Fei Ma, Qi Tian.

Collaborative multi-agent reasoning with retrieval-augmented generation for multimodal emotion recognition.

Paper Project
Overall framework of HyperEmo-RAG with hierarchical hyperbolic retrieval and structured evidence injectionHyperEmo-RAG
arXiv preprint2026

Navigating the Emotion Tree: Hierarchical Hyperbolic RAG for Multimodal Emotion Recognition

Zeheng Wang, Bo Zhao, Yijie Zhu, Zhishu Liu, Hui Ma, Ruixin Zhang, Shouhong Ding, Qianyu Xie, Zitong Yu.

Hierarchical hyperbolic retrieval and structured evidence injection for multimodal emotion recognition.

Paper
Overall framework of Atsuko with spectral-band routing and marginal complementarity supervisionAtsuko
arXiv preprint2026

Complementarity-Supervised Spectral-Band Routing for Multimodal Emotion Recognition

Zhexian Huang, Bo Zhao, Hui Ma, Zhishu Liu, Jie Zhang, Ruixin Zhang, Shouhong Ding, Zitong Yu.

Complementarity-supervised routing across spectral bands and modalities for fine-grained emotion recognition.

Paper Project

Along the way

Education & milestones

Education

  1. 2025.09 – 2027.06 (Expected)

    City University of Hong Kong (Dongguan)

    Master in Computer Science

  2. 2025.05 – 2027.08 (Expected)

    Great Bay University

    Visiting Student · YU Vision Group

  3. 2021.09 – 2025.06

    Jilin University

    Bachelor in Artificial Intelligence

Honors & awards

  • 2025CCF Dongguan Student Member Service Appreciation Award
  • 2025Jilin University Outstanding Student Leader Award
  • 2025Jilin University Third-Class Scholarship
  • 2024Outstanding Communist Youth League Cadre of Jilin University
  • 2023Second Prize · National Undergraduate Mathematical Modeling Competition
  • 2023Jilin University “Dongrong” Social Scholarship (Korea IB Group)

Hello from around the world

Visitors, assemble!

A little worldwide adventure

Every visiting region gets a Poké Ball. Thanks for stopping by!

—visits · past 30 days
—regions
World visitor map Generalized world land outlines for an interactive country-level visitor map. Country markers show approximate label locations, not a visitor's precise location. United Arab Emirates Afghanistan Albania Armenia Angola Antarctica Argentina Austria Australia Azerbaijan Bosnia and Herzegovina Bangladesh Belgium Burkina Faso Bulgaria Burundi Benin Brunei Bolivia Brazil The Bahamas Bhutan Botswana Belarus Belize Canada Democratic Republic of the Congo Central African Republic Republic of the Congo Switzerland Ivory Coast Chile Cameroon People's Republic of China Colombia Costa Rica Cuba Cyprus Czech Republic Germany Djibouti Denmark Dominican Republic Algeria Ecuador Estonia Egypt Western Sahara Eritrea Spain Ethiopia Finland Fiji Falkland Islands France Gabon United Kingdom Georgia Ghana Greenland The Gambia Guinea Equatorial Guinea Greece Guatemala Guinea-Bissau Guyana Honduras Croatia Haiti Hungary Indonesia Ireland Israel India Iraq Iran Iceland Italy Jamaica Jordan Japan Kenya Kyrgyzstan Cambodia North Korea South Korea Kuwait Kazakhstan Laos Lebanon Sri Lanka Liberia Lesotho Lithuania Luxembourg Latvia Libya Morocco Moldova Montenegro Madagascar North Macedonia Mali Myanmar Mongolia Mauritania Malawi Mexico Malaysia Mozambique Namibia New Caledonia Niger Nigeria Nicaragua Netherlands Norway Nepal New Zealand Oman Panama Peru Papua New Guinea Philippines Pakistan Poland Puerto Rico Palestine Portugal Paraguay Qatar Romania Serbia Russia Rwanda Saudi Arabia Solomon Islands Sudan Sweden Slovenia Slovakia Sierra Leone Senegal Somalia Suriname South Sudan El Salvador Syria Eswatini Chad French Southern and Antarctic Lands Togo Thailand Tajikistan East Timor Turkmenistan Tunisia Turkey Trinidad and Tobago Taiwan Tanzania Ukraine Uganda United States of America Uruguay Uzbekistan Venezuela Vietnam Vanuatu Kosovo Yemen South Africa Zambia Zimbabwe

Select a Poké Ball for visit counts. Tap the map to toss one for fun.

Loading visitor regions…

Approximate country and region locations · Visitor data by FeedPulse · Map by Natural Earth

A tiny happy break

Leave a little love.

Tap either friend to send a like. One like per visit, but as many playful taps as you wish!

—little cheers

Two tiny friends. One big thank-you.

Say hello

Let's connect.

Find my research, code, or get in touch.

EmailGitHubGoogle Scholar