Jiaxin Huang

3D/4D world models · spatial intelligence · embodied AI

Jiaxin Huang黄佳欣

Third-year Ph.D. student in Machine Learning at MBZUAI, advised by Tongliang Liu and Steve Liu, and honored to collaborate with Prof. Mingming Gong. Happy to lead the Embodied Navigation Team at AgiBot X-Lab as Project Leader, mentored by Dr. Shaoli Huang, and to work with two humanoid robots, Lingxi X2 and A3 Ultra.

My research focuses on 3D/4D world models, which I study from four complementary perspectives: spatial reasoning, 3D/4D reconstruction, embodied navigation, and the safety of 3D/4D generative models.

Always open to research discussions and collaborations. Feel free to reach out.

Seen across, a ridge; from the side, a peak. Far or near, high or low, it is never the same. I cannot see the true face of Mount Lu only because I stand inside the mountain.

Su Shi 苏轼, “Written on the Wall of Xilin Temple” 《题西林壁》, 1084
横看成岭侧成峰，远近高低各不同，不识庐山真面目，只缘身在此山中。

One world, four perspectives

A 3D/4D world model must reason about space, reconstruct dynamic scenes, support embodied action, and be protected against misuse.

The four sections below address these requirements in turn: spatial reasoning, 3D/4D reconstruction, embodied navigation and safety. Each opens a page with the papers, figures, posters and demos.

“All men by nature desire to know.”Aristotle, Metaphysics I.1 (tr. W. D. Ross)

01Spatial reasoning空间推理

How can a model ground language in 3D space, and from whose viewpoint?

Most 3D vision-language benchmarks name the target object, so a model can succeed by matching names instead of reasoning about space. I build benchmarks that remove this shortcut and methods that bring the reasoning of 2D multimodal LLMs into 3D without any 3D annotation.

子非鱼，安知鱼之乐Zhuangzi · Aristotle
East · 东

“You are not a fish. How do you know what fish enjoy?” “You are not me. How do you know I don't know what fish enjoy?”

Zhuangzi 庄子, “Autumn Floods” 《秋水》
子非鱼，安知鱼之乐，子非我，安知我不知鱼之乐。
West · 西

The soul never thinks without a mental image.

Aristotle, De Anima III.7, 431a
Open this perspective All papers in this perspective

023D/4D reconstruction三维/四维重建

What in a scene changes over time, what stays, and how to reconstruct?

A dynamic scene mixes structure that stays fixed with content that moves. I develop Gaussian splatting methods that keep the two apart, from semantic-guided reconstruction of urban scenes to physically grounded world models that turn multi-view video into simulation-ready environments.

自其变者而观之Su Shi · Aristotle
East · 东

Seen through what changes, heaven and earth do not last the blink of an eye; seen through what does not change, things and I are both without end.

Su Shi 苏轼, “Former Ode on the Red Cliff” 《赤壁赋》, 1082
盖将自其变者而观之，则天地曾不能以一瞬，自其不变者而观之，则物与我皆无尽也。
West · 西

For this is what time is: the number of motion with respect to before and after.

Aristotle, Physics IV.11, 219b
Open this perspective All papers in this perspective

04Safety of 3D/4D generative models生成安全

As videos become reconstructable 4D assets, how can the content they capture be protected?

3D Gaussian splatting and feed-forward 4D reconstruction make captured content easy to reconstruct, fine-tune and redistribute. I study proactive protection across the asset life cycle: source videos, 4D Gaussian assets, and the 3D generative models built from them.

为之于未有Laozi · Aristotle
East · 东

Act on things before they take shape; bring order before disorder begins.

Laozi 老子, Dao De Jing 《道德经》, chapter 64
为之于未有，治之于未乱。
West · 西

For man, when perfected, is the best of animals, but, when separated from law and justice, he is the worst of all.

Aristotle, Politics I.2, 1253a (tr. Benjamin Jowett)
Open this perspective All papers in this perspective

Publications

Showing 34 of 34

2026

  1. C11

    Siming-Bench: Evaluating Procedural Correctness from Continuous Interactions in Clinical Skill Videos

    Xiyang Huang, Jiawei Lin, Keying Wu, Jiaxin Huang, Kailai Yang, Renxiong Wei, Jiayi Xiang, Ziyan Kuang, Min Peng, Qianqian Xie, Sophia Ananiadou

    Multimodal & language EMNLP 2026 Oral

  2. C10

    TRACE: An Experiential Framework for Coherent Multi-hop Knowledge Graph Question Answering

    Yingxu Wang, Jiaxin Huang, Mengzhu Wang, Nan Yin

    Multimodal & language ACL 2026

  3. C9

    HUGE-Bench: A Benchmark for High-Level UAV Vision-Language-Action Tasks

    Jingyu Guo, Ziye Chen, Ziwen Li, Zhengqing Gao, Jiaxin Huang, Hanlue Zhang, Fengming Huang, Yu Yao, Tongliang Liu, Mingming Gong

    Embodied navigation ECCV 2026 Project page

    Teaser of HUGE-Bench
    TeaserHigh-level drone tasks such as building inspection, mapping and traversal, set in four real-world digital-twin scenes with a collision-aware simulator.
  4. C8

    Mobile-VTON: High-Fidelity On-Device Virtual Try-On

    Zhenchen Wan, Ce Chen, Runqi Lin, Jiaxin Huang, Tianxi Chen, Yanwu Xu, Tongliang Liu, Mingming Gong

    Multimodal & language CVPR 2026

  5. J7

    Joint Lossless Compression and Steganography for Medical Images via Large Language Models

    Pengcheng Zheng, Xiaorong Pu, Kecheng Chen, Jiaxin Huang, Meng Yang, Bai Feng, Yazhou Ren, Jianan Jiang, Chaoning Zhang, Yang Yang, Heng Tao Shen

    Medical & industrial AI IEEE TMM 2026

  6. S16

    GuidedNav: Shaping Vision-Language Navigation Representations through Pre-Action Attention and Spatial Guidance

    Jingyu Guo, Jiaxin Huang†, Chengxing Lin, Xiaoyu Luo, Cheng Wen, Shanshan Ye, Tingjin Chu, Shaoli Huang, Mingming Gong

    Embodied navigation arXiv 2026 Under review Project page

    Teaser & framework of GuidedNav
    TeaserThree kinds of guidance along one route: which subtask comes next, which landmark to attend to, and how to move in 3D.
    FrameworkGuidance acts at the pre-action position and only during training: subtask attention over the instruction, landmark attention over the current frame, and long-horizon and local spatial guidance. At inference the policy runs without these modules.
  7. S15

    NavTTT: Adaptive Parametric Memory via Hierarchical Test-Time Training for Vision-and-Language Navigation

    Xiaoyu Luo, Chengxing Lin, Jiaxin Huang†, Jiale Liu, Jingyu Guo, Sirui Chen, Aolin Li, Shaoli Huang, Cheng-zhong Xu, Hui Kong

    Embodied navigation Under review

  8. S14

    EgoAudit: Auditing Egocentric Spatial Reasoning in 3D Vision-Language Models

    Jiaxin Huang, Chengcheng Lin, Xiyang Huang, Yankai Chen, Min Peng, Qianqian Xie

    Spatial reasoning Under review

  9. S13

    Safe4D: Causal Video Protection against Unauthorized 4D Reconstruction

    Jiaxin Huang, Ziming Hong, Mingming Gong, Tongliang Liu

    Safety Under review

  10. S12

    4DGuard: Towards Proactive 4D Gaussian Assets Protection for Dynamic Novel-View Synthesis

    Jiaxin Huang, Ziming Hong

    Safety Under review

  11. S11

    Position: Interaction Understanding Requires More Than Object-Centric Representations

    Jiaxin Huang, Hai Huang

    Multimodal & language Under review

  12. S10

    Discrete Diffusion Models: A Unified Framework from Tokenization to Generation

    Ye Yuan, Weien Li, Rui Song, Zeyu Li, Haochen Liu, Xiangyu Kong, Zixuan Dong, Linfeng Du, Zipeng Sun, Weixu Zhang, Jiaxin Huang, Changjiang Han, Yonghan Yang, Zichen Zhao, Xiuyuan Hu, Haolun Wu, Yankai Chen, Fengran Mo, Jikun Kang, Bowei He, Dawn Song, Philip S. Yu, Xue Liu

    Multimodal & language arXiv 2026

  13. S9

    Predicting the World via Video Representation: A Comprehensive Survey on Video World Models

    Jiaxin Yan, Chaoning Zhang, Xudong Wang, Pengcheng Zheng, Ya Wen, Qigan Sun, Jiaxin Huang, Shuxu Chen, Yang Yang, Hyundong Shin

    3D/4D reconstruction Preprints 2026

  14. S8

    Immunizing 3D Gaussian Generative Models Against Unauthorized Fine-Tuning via Attribute-Space Traps

    Jianwei Zhang, Sihan Cao, Chaoning Zhang, Ziming Hong, Jiaxin Huang, Pengcheng Zheng, Caiyan Qin, Wei Dong, Yang Yang, Tongliang Liu

    Safety arXiv 2026

  15. S7

    RCP: Representation Consistency Pruner for Mitigating Distribution Shift in Large Vision-Language Models

    Jianwei Zhang, Chaoning Zhang, Sihan Cao, Wang Liu, Pengcheng Zheng, Jiaxin Huang, Caiyan Qin, Yalan Ye, Wei Dong, Yang Yang

    Multimodal & language arXiv 2026

  16. S6

    DSBD: Dual-Aligned Structural Basis Distillation for Graph Domain Adaptation

    Yingxu Wang, Kunyu Zhang, Jiaxin Huang, Mengzhu Wang, Mingyan Xiao, Siyang Gao, Nan Yin

    Multimodal & language arXiv 2026

  17. S5

    ExecVLA: Following Fine-Grained Execution Constraints in Vision-Language-Action Models with Bi-Level Action Representation

    Gaoge Han, Zhengqing Gao, Ziwen Li, Jiaxin Huang, Shaoli Huang, Fakhri Karray, Mingming Gong, Tongliang Liu

    Embodied navigation arXiv 2026

  18. S4

    Intellectual Property Protection for 3D Gaussian Splatting Assets: A Survey

    Longjie Zhao, Ziming Hong, Jiaxin Huang, Runnan Chen, Mingming Gong, Tongliang Liu

    Safety arXiv 2026

  19. S3

    Towards Visual Chain-of-Thought Reasoning: A Comprehensive Survey

    Pengcheng Zheng, Chaoning Zhang, Mingzheng Cui, Guo Chen, Qigan Sun, Jiaxin Huang, Jiaquan Zhang, Tae-Ho Kim, Caiyan Qin, Yazhou Ren, Guoqing Wang, Yang Yang

    Multimodal & language TechRxiv 2026

  20. S2

    Mirage2Matter: A Physically Grounded Gaussian World Model from Video

    Zhengqing Gao, Ziwen Li, Xin Wang, Jiaxin Huang, Zhenyang Ren, Mingkai Shao, Hanlue Zhang, Tianyu Huang, Yongkang Cheng, Yandong Guo, Runqi Lin, Yuanyuan Wang, Tongliang Liu, Kun Zhang, Mingming Gong

    3D/4D reconstruction arXiv 2026

2025

  1. C7

    SURPRISE3D: A Dataset for Spatial Understanding and Reasoning in Complex 3D Scenes

    Jiaxin Huang*, Ziwen Li*, Hanlue Zhang*, Runnan Chen, Zhengqing Gao, Xiao He, Yandong Guo, Wenping Wang, Tongliang Liu, Mingming Gong

    Spatial reasoning NeurIPS 2025, Datasets & Benchmarks Track Proceedings

    Teaser of SURPRISE3D
    TeaserSURPRISE3D asks for objects without naming them. Spatial queries cover relative position, absolute distance, and narrative or parameter perspectives; knowledge queries cover common sense and human intention.
  2. C6

    MLLM-For3D: Adapting Multimodal Large Language Model for 3D Reasoning Segmentation

    Jiaxin Huang, Runnan Chen, Ziwen Li, Zhengqing Gao, Xiao He, Yandong Guo, Mingming Gong, Tongliang Liu

    Spatial reasoning NeurIPS 2025 Proceedings

    Framework of MLLM-For3D
    FrameworkA frozen 2D multimodal LLM segments several posed views from one implicit query. Token attention keeps the reliable views, and their masks become pseudo-labels that train a 3D encoder without any 3D labels.
  3. C5

    Large Language Models for Lossless Image Compression: Next-Pixel Prediction in Language Space is All You Need

    Kecheng Chen, Pingping Zhang, Hui Liu, Jie Liu, Yibing Liu, Jiaxin Huang, Shiqi Wang, Hong Yan, Haoliang Li

    Multimodal & language NeurIPS 2025

  4. C4

    Lightweight Medical Image Restoration via Integrating Reliable Lesion-Semantic Driven Prior

    Pengcheng Zheng, Kecheng Chen, Jiaxin Huang, Bohao Chen, Ju Liu, Yazhou Ren, Xiaorong Pu

    Medical & industrial AI ACM MM 2025

  5. J6

    ProtoMol: Enhancing Molecular Property Prediction via Prototype-Guided Multimodal Learning

    Yingxu Wang, Kunyu Zhang, Jiaxin Huang, Nan Yin, Siwei Liu, Eran Segal

    Multimodal & language Brief. Bioinform. 2025

  6. J5

    Data-Driven Soft Fault Diagnosis for Analog Circuits Based on Contrastive Learning

    Naixin Zhou*, Jiaxin Huang*, Yijiu Zhao, Shibo Chen, Jiaming Zhao, Feiyu Long

    Medical & industrial AI IEEE TIM 2025

2024

  1. C3

    Sparse Bayesian Deep Learning for Cross Domain Medical Image Reconstruction

    Jiaxin Huang, Qi Wu, Yazhou Ren, Fan Yang, Aodi Yang, Qianqian Yang, Xiaorong Pu

    Medical & industrial AI AAAI 2024

  2. J4

    Cross-Domain Low-Dose CT Image Denoising with Semantic Preservation and Noise Alignment

    Jiaxin Huang, Kecheng Chen, Yazhou Ren, Jiayu Sun, Xiaorong Pu, Ce Zhu

    Medical & industrial AI IEEE TMM 2024

  3. S1

    UrbanGS: Semantic-Guided Gaussian Splatting for Urban Scene Reconstruction

    Ziwen Li*, Jiaxin Huang*, Runnan Chen, Yunlong Che, Yandong Guo, Tongliang Liu, Fakhri Karray, Mingming Gong

    3D/4D reconstruction arXiv 2024

2023

  1. J3

    Blockchain-Based Interpretable Electric Vehicle Battery Life Prediction in IoV

    Siyan Liu*, Chang Wu*, Jiaxin Huang*, Ying Zhang, Ming Ye, Yuhang Huang

    Medical & industrial AI IEEE IoT-J 2023

  2. J2

    CDDnet: Cross-Domain Denoising Network for Low-Dose CT Image via Local and Global Information Alignment

    Jiaxin Huang, Kecheng Chen, Yazhou Ren, Jiayu Sun, Yanmei Wang, Tao Tao, Xiaorong Pu

    Medical & industrial AI Comput. Biol. Med. 2023

2022

  1. C2

    Cross Domain Low-Dose CT Image Denoising with Semantic Information Alignment

    Jiaxin Huang, Kecheng Chen, Jiayu Sun, Xiaorong Pu, Yazhou Ren

    Medical & industrial AI ICIP 2022 Oral

  2. J1

    A Survey on Clinical Oriented CT Image Denoising

    Xiaorong Pu*, Jiaxin Huang*, Junchi Liu, Jiayu Sun, Jixiang Luo, Yue Zhao, Kecheng Chen, Yazhou Ren

    Medical & industrial AI Front. Data Comput. 2022

2020

  1. C1

    Task-Driven Deep Learning for LDCT Image Denoising

    Kecheng Chen*, Jiaxin Huang*, Jiayu Sun, Yazhou Ren, Xiaorong Pu

    Medical & industrial AI ISICDM 2020

* equal contribution · † project leader

News

Always up for a chat about world models, embodied AI and spatial intelligence (or HYROX and specialty coffee). Drop me an email!

  1. GuidedNav is live on arXiv, with a project page full of simulation and real-robot rollouts! Huge thanks to our navigation team at AgiBot X-Lab. Leading this project has been a blast.

  2. Co-organized LifeGenIP at ECCV 2026 in Malmö, a workshop on protecting intellectual property across the whole life cycle of visual generative models. Thank you to every speaker and participant!

  3. Six papers submitted, four as first author and two as project leader. Fingers crossed!

  4. Siming-Bench, led by my mentee Xiyang Huang, is an oral paper at EMNLP 2026. So proud of him!

  5. A great year with collaborators: HUGE-Bench at ECCV, TRACE at ACL and Mobile-VTON at CVPR.

  6. Joined AgiBot X-Lab in Shenzhen to lead the Embodied Navigation Team, teaching humanoid robots to find their way.

  7. Teaching assistant for ML813, Dimensionality Reduction and Manifold Learning, a Ph.D. course at MBZUAI.

  8. Two papers at NeurIPS 2025: SURPRISE3D and MLLM-For3D, both on 3D spatial reasoning.

  9. Teaching assistant for Linear Algebra and Calculus for MBZUAI's very first cohort of AI undergraduates.

  10. Started my Ph.D. in Machine Learning at MBZUAI in Abu Dhabi.

About

Education and industry

  1. 2024 to presentPh.D., Machine LearningMBZUAI, Abu Dhabi. Advisors: Prof. Tongliang Liu and Prof. Steve Liu.
  2. Apr to Oct 2026Research Intern; Project Leader, Embodied Navigation TeamAgiBot X-Lab, Shenzhen. Mentors: Dr. Shaoli Huang and Zhihui Peng (co-founder of AgiBot).
  3. 2021 to 2024Master's, Computer Science and TechnologyUniversity of Electronic Science and Technology of China (UESTC), Chengdu.
  4. 2017 to 2021B.Sc., AutomationUESTC. GPA 3.89/4.00; National Scholarship, ranked 1st of 297.
  5. VisitsVisiting student, The University of SydneySummer session, University of California, Berkeley.

Teaching and mentoring

  • ML813 Dimensionality Reduction and Manifold Learning, Ph.D. course, MBZUAI, Spring 2026. Teaching assistant.
  • Linear Algebra and Calculus, first cohort of AI undergraduates, MBZUAI, 2025. Teaching assistant.
  • Theory and Application of Neural Networks, UESTC, 2022. Teaching assistant.
  • Mentored the AgiBot navigation interns Xiaoyu Luo, Jingyu Guo, Aolin Li, Chengxing Lin and Sirui Chen (GuidedNav, NavTTT), and Xiyang Huang of Wuhan University (EMNLP 2026 oral), along with students at UESTC, MBZUAI and McGill.

Service and honors

  • Co-organizer, LifeGenIP workshop, ECCV 2026.
  • Reviewer for CVPR, ECCV and ICCV (2024 to 2026), NeurIPS (2025 to 2026), ICLR and ICML (2025 to 2026), and AAAI 2025.
  • National Scholarship, ranked 1st of 297, Ministry of Education of China, 2020.
  • First Prize, ISICDM 2020 challenge on low-dose CT image quality.
  • “Liren” Elite of UESTC, one of five students university-wide, 2018.

Off the lab · 生活

Life outside research

When I look at things through myself, everything takes on my colours.

Wang Guowei 王国维, Remarks on Song Ci Poetry 《人间词话》, 1908
以我观物，故物皆著我之色彩。

Move

  • HYROX
  • Spartan Race
  • CrossFit
  • Hiking

Fitness racing, obstacle racing, CrossFit training, and hiking whenever there is a trail.

Row hard, row hard, and a whole sandbar of gulls and egrets takes flight.

Li Qingzhao 李清照, to the tune “Like a Dream” 《如梦令》
争渡，争渡，惊起一滩鸥鹭。

Taste

  • Specialty coffee
  • Chinese tea

Good coffee beans and Chinese tea, and the quiet that comes with both.静水流深 · 静能生慧Still water runs deep; stillness brings insight.

Evening falls and snow is on the way. Will you share a cup with me?

Bai Juyi 白居易, “A Question for Liu Nineteen” 《问刘十九》
晚来天欲雪，能饮一杯无。

Read

  • Anne Carson
  • Classical poetry

Anne Carson, the Canadian poet and classicist, received the 2026 Nobel Prize in Literature for work “in playful dialogue with the classical tradition”. Three of her ideas I keep returning to: the best art carries you along with it; time is not a thing at all but a sense we read into movement; and a sentence half-dreamed at night can open a door to somewhere else.

Open full size