Max W.F. Ku

I am a third-year PhD student in Computer Science at the University of Waterloo, Faculty of Mathematics, advised by Wenhu Chen and Yuntian Deng. Previously, I interned at Adobe Research and NVIDIA Research.

My research focuses on understanding and editing worlds, spanning controllable generation, evaluation, physical reasoning, and world editing.

I am especially interested in how generative models can make precise interventions while preserving the structure and dynamics of the world. My current research interests include

  • World Editing: making precise interventions while preserving the structure and dynamics of the world.
  • Controllable Video and Image Editing: precise, counterfactual, and temporally consistent manipulation.
  • World and Physical Reasoning: understanding dynamics, causality, and how changes propagate through a scene.
  • Evaluation and Interpretability: grounded and explainable evaluation of generative models.

Professional Activities

  • Reviewer for Conferences & Journals: ICLR, NeurIPS, ICML, SIGGRAPH Asia, SIGGRAPH, TVCG, TIP, ACL, EMNLP, TMLR
  • Reviewer for Workshops: ICMLW, ECCVW, WACVW, NeurIPSW

Community

  • I lead GGG, a community-driven group dedicated to sharing and discussing papers on Generative AI.
  • I host Online Coffee Chat to share advice with students from underrepresented backgrounds.

Misc

news

Sep 26, 2026 My work at NVIDIA (PhyProbe) is Accepted to NeurIPS 2026.
Jun 22, 2026 Joined Adobe Research as an intern for Summer 2026.
Jan 25, 2026 ImagenWorld and EditReward Accepted to ICLR 2026.
Jun 15, 2025 Achieved a total of 1000 citations.
Jun 02, 2025 Joined NVIDIA Deep Imagination Research as an intern for Summer 2025.

selected publications

  1. Preprint
    igmworld.png
    World Editing: Intervening on Executable Worlds at Increasing Depth
    Max Ku, Nok-Kan Law, Yu-Chien Tang, Shih-Ying Yeh, Ping Nie, Andy Zheng, Tat Hei Lai, Fei-Yueh Chen, Nikko Yu, Wei-Chieh Sun, Suzy Huang, Chiao-Wei Hsu, Chih-Chuan Huang, Chak-Wing Mak, Ho Yin Sam Ng, Edisy Kin Wai Chan, Min-Hung Chen, and Ho Kei Cheng
    In arXiv , 2026
  2. Preprint
    viescore2.png
    VIEScore2: Unified Image Evaluation with Spatially Grounded Explanations
    Xianda Du*, Max Ku*, Weiming Ren, Zhi Rui Tam, Chunlin Ren, Ping Nie, Min-Hung Chen, and Wenhu Chen
    In arXiv , 2026
  3. NeurIPS 2026
    phyprobe.png
    PhyProbe: Rethinking Physical Consistency Evaluation in Generated Videos
    Max Ku, Jiaojiao Fan, Zekun Hao, Francesco Ferroni, Heng Wang, Wenhu Chen, Ming-Yu Liu, and Prithvijit Chattopadhyay
    In The Fortieth Annual Conference on Neural Information Processing Systems , 2026
  4. Preprint
    visphyworld.png
    VisPhyWorld: Probing Physical Reasoning via Code-Driven Video Reconstruction
    Jiarong Liang*, Max Ku*, Ka-Hei Hui, Ping Nie, and Wenhu Chen
    In arXiv , 2026
  5. ICLR 2026
    imagenworld.png
    ImagenWorld: Stress-Testing Image Generation Models with Explainable Human Evaluation on Open-ended Real-World Tasks
    Samin Mahdizadeh Sani*, Max Ku*, Nima Jamali, Matina Mahdizadeh Sani, Paria Khoshtab, Wei-Chieh Sun, Parnian Fazel, Zhi Rui Tam, Thomas Chong, Edisy Kin Wai Chan, Donald Wai Tong Tsang, Chiao-Wei Hsu, Ting Wai Lam, Ho Yin Sam Ng, Chiafeng Chu, Chak-Wing Mak, Keming Wu, Hiu Tung Wong, Yik Chun Ho, Chi Ruan, Zhuofeng Li, I-Sheng Fang, Shih-Ying Yeh, Ho Kei Cheng, Ping Nie, and 1 more author
    In The 14th International Conference on Learning Representations , 2026
  6. ACL 2025 Oral
    tha.png
    TheoremExplainAgent: Towards Multimodal Explanations for LLM Theorem Understanding
    Max Ku*, Thomas Chong*, Jonathan Leung, Krish Shah, Alvin Yu, and Wenhu Chen
    In The 63rd Annual Meeting of the Association for Computational Linguistics , 2025
  7. NeurIPS 2024
    genai_arena.png
    GenAI Arena: An Open Evaluation Platform for Generative Models
    Dongfu Jiang*, Max Ku*, Tianle Li*, Yuansheng Ni, Shizhuo Sun, Rongqi Fan, and Wenhu Chen
    In The Conference on Neural Information Processing Systems , 2024
  8. TMLR 2024
    anyv2v.png
    AnyV2V: A Tuning-Free Framework For Any Video-to-Video Editing Tasks
    Max Ku*, Cong Wei*, Weiming Ren*, Harry Yang, and Wenhu Chen
    Transactions on Machine Learning Research, 2024
  9. ACL 2024
    viescore.png
    VIEScore: Towards Explainable Metrics for Conditional Image Synthesis Evaluation
    Max Ku, Dongfu Jiang, Cong Wei, Xiang Yue, and Wenhu Chen
    In The 62nd Annual Meeting of the Association for Computational Linguistics , 2024