Yikai Wang

About

Yikai Wang is a postdoctoral researcher at Meta, London, working with Prof. Tao Xiang. Prior to this, he was a research fellow at MMLab@NTU, Nanyang Technological University, Singapore, working with Prof. Chen Change Loy. He received his Ph.D. in Statistics and B.Sc. in Mathematics from Fudan University, Shanghai, advised by Prof. Yanwei Fu.

His research focuses on structured generative models that infer and represent both observed and unobserved aspects of the visual world. His work has led to several patents and proprietary models that have been acquired and deployed in commercial applications. He serves as an area chair and reviewer for major conferences and journals, and has co-organized several workshops and tutorials.

At Meta, he works on training video generative models.

Reach out via yi-kai.wang@outlook.comcopied ✓. Please let me know who you are.
My previous institutional emails from NTU and Fudan are no longer active.

Yikai Wang reconstructed from coarse to fine with Pixel-NVG
Pixel-NVG Reconstruction

News

Research

Papers

In-Flight KV Cache with Clean Anchors for Faster Autoregressive Video Diffusion.
[arXiv] [project page] [video intro: YouTube, Bilibili]
Yikai Wang, Xiao Han, Mengmeng Xu, Juan C. Pérez, Yiannis Douratsos, Sen He, Zijian Zhou, Fei Zhang, Zhaochong An, Juan-Manuel Pérez-Rúa, Chen Change Loy, Tao Xiang.
Preprint, 2026.

AffineTok: Semantic Affine Consistency for Diffusion-Friendly Visual Tokenizer.
[arXiv] [project page]
Junqiu Yu, Pandeng Li, Yikai Wang, Jiaxing Zhao, Yujie Wei, Kaixun Jiang, Quanhao Li, Hongtao Yu, Zhihang Liu, Zhaohe Liao, Junjie Zhou, Yun Zheng, Yu Liu, Yanwei Fu.
Preprint, 2026.

World Action Modeling with Progressive Visual Planning.
[arXiv] [project page]
Fei Zhang, Zhaochong An, Duncan Frost, Yikai Wang, Pengfei Liu, Ya Zhang, Michal Drozdzal, Amir Bar.
Preprint, 2026.

Native Action-Prior Learning from Videos for World Action Models.
[arXiv] [project page]
Zhaochong An, Fei Zhang, Menglin Jia, Duncan Frost, Zijian Zhou, Yikai Wang, Xudong Wang, Aditya Patel, Belinda Zeng, Tao Xiang, Serge Belongie, Amir Bar, Sen He.
Preprint, 2026.

Aligned Stable Inpainting: Mitigating Unwanted Object Insertion and Preserving Color Consistency.
[arXiv]
Yikai Wang*, Junqiu Yu*, Chenjie Cao, Xiangyang Xue, Yanwei Fu.
Preprint, 2026.

The Pictorial Cortex: Zero-Shot Cross-Subject fMRI-to-Image Reconstruction via Compositional Latent Modeling.
[arXiv]
Jingyang Huo, Yikai Wang, Yanwei Fu, Jianfeng Feng.
Preprint, 2026.

Direct 3D-Aware Object Insertion via Decomposed Visual Proxies.
[arXiv] [paper] [code] [project page]
Jingbo Gong, Yikai Wang, Yushi Lan, Yuhao Wan, Ziheng Ouyang, Rui Zhao, Ming-Ming Cheng, Qibin Hou, Chen Change Loy.
ICML, 2026.

Conformal Reliability: A New Evaluation Metric for Conditional Generation.
[arXiv] [paper] [code] [project page]
Yachen Gao, Xinwei Sun, Yikai Wang, Ye Shi, Jingya Wang, Jianfeng Feng, Yanwei Fu.
ICML, 2026.

ActiveVLA: Injecting Active Perception into Vision-Language-Action Models for Precise 3D Robotic Manipulation.
[arXiv] [paper] [code] [data] [project page]
Zhenyang Liu, Yongchong Gu, Yikai Wang, Xiangyang Xue, Yanwei Fu.
CVPR, 2026.

Next Visual Granularity Generation.
[arXiv] [paper] [code] [project page] [简介]
Yikai Wang, Zhouxia Wang, Zhonghua Wu, Qingyi Tao, Kang Liao, Chen Change Loy.
ICLR, 2026.

Thinking with Camera: A Unified Multimodal Model for Camera-Centric Understanding and Generation.
[arXiv] [paper] [code] [project page] [简介]
Kang Liao, Size Wu, Zhonghua Wu, Linyi Jin, Chao Wang, Yikai Wang, Fei Wang, Wei Li, Chen Change Loy.
ICLR, 2026.

fMRI2Face: Geometry-Guided Neural Decoding Framework and a Full-HD fMRI-Video Dataset for Controllable Digital Human Faces.
[arXiv]
Jingyang Huo, Xiangru Huang, Chentao Shen, Yikai Wang, Yun Wang, Jianxiong Gao, Yulei Xin, Shihao Jin, Yanwei Fu, Jianfeng Feng.
Neurocomputing, 2026.

Cross-subject fMRI-to-Image with Visual-cortex 2D Representation and Pre-Training.
[paper]
Xuelin Qian*, Yun Wang*, Yikai Wang, Shijie Zhao, Jianfeng Feng, Yanwei Fu.
JBHI, 2026.

Spatial-Temporal Aware Visuomotor Diffusion Policy Learning.
[arXiv] [paper] [project page]
Zhenyang Liu, Yikai Wang, Kuanning Wang, Longfei Liang, Xiangyang Xue, Yanwei Fu.
ICCV, 2025.

Towards Enhanced Image Inpainting: Mitigating Unwanted Object Insertion and Preserving Color Consistency.
[arXiv] [paper] [full-size PDF] [code & MISATO dataset] [project page] [demo: YouTube, Bilibili]
Yikai Wang*, Chenjie Cao*, Junqiu Yu*, Ke Fan, Xiangyang Xue, Yanwei Fu.
CVPR, 2025. (Highlight, 13.5% of accepted papers)

ReasonGrounder: LVLM-Guided Hierarchical Feature Splatting for Open-Vocabulary 3D Visual Grounding and Reasoning.
[arXiv] [paper] [code & dataset] [project page]
Zhenyang Liu, Yikai Wang, Sixiao Zheng, Tongying Pan, Longfei Liang, Yanwei Fu, Xiangyang Xue.
CVPR, 2025.

Adaptive Pruning of Pretrained Transformer via Differential Inclusions.
[arXiv] [paper] [code]
Yizhuo Ding, Ke Fan, Yikai Wang, Xinwei Sun, Yanwei Fu.
ICLR, 2025.

3D StreetUnveiler with Semantic-aware 2DGS - a simple baseline.
[arXiv] [paper] [code] [project page]
Jingwei Xu, Yikai Wang, Yiqun Zhao, Yanwei Fu, Shenghua Gao.
ICLR, 2025.

Repositioning the Subject within Image.
[arXiv] [paper] [full-size PDF] [code & ReS dataset] [project page] [demo: YouTube, Bilibili]
Yikai Wang, Chenjie Cao, Ke Fan, Qiaole Dong, Yifan Li, Xiangyang Xue, Yanwei Fu.
TMLR, 2024. (J2C Certification, 10% of accepted papers; Poster at ICML 2026)

LEA: Learning Latent Embedding Alignment Model for fMRI Decoding and Encoding.
[arXiv] [paper] [code]
Xuelin Qian*, Yikai Wang*, Xinwei Sun, Yanwei Fu, Xiangyang Xue, Jianfeng Feng.
TMLR, 2024.

Unified Lexical Representation for Interpretable Visual-Language Alignment.
[arXiv] [paper] [code] [project page]
Yifan Li, Yikai Wang, Yanwei Fu, Dongyu Ru, Zheng Zhang, Tong He.
NeurIPS, 2024.

Towards Global Optimal Visual In-Context Learning Prompt Selection.
[arXiv] [paper] [code] [project page]
Chengming Xu*, Chen Liu*, Yikai Wang, Yuan Yao, Yanwei Fu.
NeurIPS, 2024.

NeuroPictor: Refining fMRI-to-Image Reconstruction via Multi-individual Pretraining and Multi-level Modulation.
[arXiv] [paper] [code] [project page]
Jingyang Huo*, Yikai Wang*, Yun Wang*, Xuelin Qian, Chong Li, Yanwei Fu, Jianfeng Feng.
ECCV, 2024.

LeftRefill: Filling Right Canvas based on Left Reference through Generalized Text-to-Image Diffusion Model.
[arXiv] [paper] [code] [project page]
Chenjie Cao, Yunuo Cai, Qiaole Dong, Yikai Wang, Yanwei Fu.
CVPR, 2024.

Test-Time Linear Out-of-Distribution Detection.
[paper] [code]
Ke Fan*, Tong Liu*, Xingyu Qiu, Yikai Wang, Lian Huai, Zeyu Shangguan, Shuang Gou, Fengjian Liu, Yuqian Fu, Yanwei Fu, Xingqun Jiang.
CVPR, 2024.

Coarse-to-Fine Amodal Segmentation with Shape Prior.
[arXiv] [paper] [code] [project page]
Jianxiong Gao, Xuelin Qian, Yikai Wang, Tianjun Xiao, Tong He, Zheng Zhang, Yanwei Fu.
ICCV, 2023.

Knockoffs-SPR: Clean Sample Selection in Learning with Noisy Labels.
[arXiv] [paper] [code] [project page] [简介]
Yikai Wang*, Yanwei Fu*, Xinwei Sun.
TPAMI, 2023.

Scalable Penalized Regression for Noise Detection in Learning with Noisy Labels.
[arXiv] [paper] [code] [project page] [简介]
Yikai Wang, Xinwei Sun, Yanwei Fu.
CVPR, 2022.

How to Trust Unlabeled Data? Instance Credibility Inference for Few-Shot Learning.
[arXiv] [paper] [code] [project page] [简介]
Yikai Wang, Li Zhang, Yuan Yao, Yanwei Fu.
TPAMI, 2021.

Instance Credibility Inference for Few-Shot Learning.
[arXiv] [paper] [code] [project page] [简介]
Yikai Wang, Chengming Xu, Chen Liu, Li Zhang, Yanwei Fu.
CVPR, 2020.

Talks

A Decade of Denoising Turned Noise into Worlds: Diffusion Models, 2015–2026.
[slides] [YouTube] [Bilibili]
Recorded talk, 2026.10.

Structured Generative Vision: From Context-Stable Inpainting to Next Visual Granularity Generation.
[slides]
Xi'an Jiaotong University, 2026.05;
Fudan University, 2026.05;
Shanghai Innovation Institute, 2026.05.

Filling Right Canvas based on Left Reference through Generalized Text-to-Image Diffusion Model.
[slides] [Bilibili video (in Chinese, starts at 1:18:40)]
Oral presentation at Pre-CVPR@Shanghai, 2024.05.

Advancing Image Inpainting: From Versatility to Consistency.
[slides]
Nanyang Technological University, 2024.04.

Clean Sample Selection Algorithms with Statistical Sparsity Analysis.
[slides]
Technical University of Munich, 2024.03;
Institute of Science and Technology Austria, 2024.04.

Few-shot Learning by Statistical Methods.
[slides]
CVPR 2023 tutorial, 2023.06:
Few-shot Learning from Meta-Learning, Statistical Understanding to Applications.

Sparse Learning for Noisy Data Detection.
[slides] [YouTube] [Bilibili]
CVPR 2022 tutorial, 2022.06:
Sparse Learning in Neural Networks and Robust Statistical Analysis.


Grants and Awards

Timeline

Service

Teaching