Zhiqi Huang

Zhiqi Huang

Modeling & Simulation for Robot Learning

Zhiqi Huang

I am a Research Assistant at The Chinese University of Hong Kong, Shenzhen and an M.Phil. student in Information Architecture at Waseda University.

My earlier research and engineering work centered on neural simulation and scalable simulation. I now focus more on simulation-driven embodied AI, particularly deformable object manipulation.

News

  • [Sep. 2026] FTC-Seg was accepted to NeurIPS 2026.
  • [Jul. 2026] Joined CUHK-Shenzhen as a Research Assistant.
  • [Jul. 2026] Two first-author manuscripts are under review.
  • [Jan. 2026] SIE3D was accepted to ICASSP 2026.

Research Interests

  • Neural Simulation: How can neural models turn observations into controllable and predictive 3D simulations?
  • Scalable Simulation: How can we build, run, and adapt simulations at scale within limited data and compute budgets?
  • Simulation-driven Embodied AI: How can we bridge the gap between simulation and reality to turn simulated experience into reliable robot skills?

Publications

When Noise Meets Long-Tail: Feature-Threshold Dual Calibration for Robust Pseudo-Labeling

NeurIPS 2026 - Second Author, Corresponding Author
Ping Guo, Zhiqi Huang*, Xinran Li

FTC-Seg tackles data-efficient visual learning under noise and class imbalance. Residual feature refinement and class-adaptive pseudo-label thresholds make better use of unlabeled observations for semi-supervised semantic segmentation, reducing reliance on manual annotation.

arXiv Code

SIE3D: Single-Image Expressive 3D Avatar Generation via Semantic Embedding and Perceptual Expression Loss

ICASSP 2026 - First Author, Corresponding Author
Zhiqi Huang*, Dulongkai Cui, Jinglu Hu

SIE3D generates expressive, controllable 3D head avatars from a single image. Text conditioning and an expression-aware loss balance identity preservation with appearance and expression control, supporting rendering from different viewpoints.

Project Page IEEE Xplore arXiv Code

Manuscripts Under Review

Controllable 3D Material Generation

First Author - Manuscript Under Review

A neural model generates controllable, relightable materials from geometry and language. Explicit reflectance properties let the same 3D asset be rendered consistently across viewpoints and lighting conditions, with appearance controlled through text.

Predictive Modeling of 3D Medical Images

First Author - Manuscript Under Review

A neural model forecasts 3D lesion evolution from longitudinal CT observations. Observed history and structural constraints guide predictions of future geometry and volumetric appearance, capturing temporal change while encouraging anatomical consistency.

Experience

The Chinese University of Hong Kong, Shenzhen

Research Assistant - Jul. 2026 - Present

Exploring simulation-driven embodied AI for deformable object manipulation, with an emphasis on learning from simulated experience and adapting through real-world interaction.

Waseda University

Research Assistant - 2024 - 2025

Developed neural methods for 3D appearance simulation, generating controllable head avatars and relightable materials from visual and semantic inputs.

4399 Games

Graphics Engineer / Senior Graphics Engineer - 2021 - 2024

Focused on scalable 3D simulation, improving computational efficiency and building reusable pipelines across platforms.

  • Appearance simulation: Developed shaders and physically based material systems for Era of Conquest, alongside reusable asset and rendering pipelines across mobile, PC, and web.
  • Computational efficiency: Optimized rendering, physics solvers, and simulation engine runtime, with a focus on real-time performance on mobile devices.
  • Engineering leadership: Promoted to Senior Graphics Engineer in 2023; led a rendering team of 3-5 engineers and coordinated the rendering roadmap across mobile and PC.

Education

  • Waseda University, Fukuoka, Japan
    • M.Phil. in Information Architecture, Apr. 2025 - Mar. 2027 (expected)
    • English-taught program; current GPA: 3.8 / 4.0
  • Sun Yat-sen University, Guangzhou, China
    • B.E. in Software Engineering, 2017 - 2021
    • GPA: 3.7 / 4.0

Honors and Scholarships

  • Partial Tuition-Waiver Scholarship for Privately Financed International Students, Waseda University, 2026-2027 (recognizing outstanding academic performance)
  • Outstanding Student Scholarship, Sun Yat-sen University, 2018-2019

Languages

  • Chinese: Native (Mandarin and Cantonese)
  • English: Professional working proficiency (TOEFL iBT: 90)