Ze Wang

I work on frontier multimodal generative foundation models — joint video, audio, and synchronized audiovisual generation at scale. At Luma, I am a primary contributor to Ray 3.5, Luma’s first joint audio‑video generation model, and to the unified Omni foundation model family.

Previously, I was a Research Scientist at AMD GenAI, where I led the development of vision generation and understanding models in the fully open-source Instella model family, including Instella‑T2I. Before that, I was a Postdoctoral Research Associate at Purdue University. I received my Ph.D. from Purdue in 2023, advised by Prof. Qiang Qiu and Prof. Guillermo Sapiro (starting at Duke University), and my B.E. from Beihang University in 2017.

My research interests include diffusion, autoregressive, and hybrid models for efficient image, video, and audio generation; multimodal large language models for unified understanding and generation; and post-training of foundation models.

Experience

• 2026 – now
  Luma AI · Research Scientist

  Joint audio‑video generation (Ray 3.5), the Omni foundation model family, and Luma’s audio and audiovisual modeling stack.

• 2024 – 2026
  AMD · Research Scientist, GenAI

  Led Instella‑T2I (text‑to‑image), Instella‑Nexus (unified multimodal), and Instella‑T2V (text‑to‑video) in the open-source Instella model family.

• 2023 – 2024
  Purdue University · Postdoctoral Research Associate

  Efficient diffusion models, generative model evaluation, and adaptation of vision and language foundation models.

• earlier
  Research internships

  Microsoft Azure AI (2022–2024), Meta AI (2021), Volvo AI Research (2020), Horizon Robotics (2017).

Selected Publications

1. Instella-T2I: Pushing the Limits of 1D Discrete Latent Space Image Representation and Generation

   Ze Wang, Hao Chen, Jiang Liu, Ximeng Sun, Jialian Wu, Yusheng Su, Xiaodong Yu, Emad Barsoum, Zicheng Liu

   preprint 2025 code

2. CD4LM: Consistency Distillation and aDaptive Decoding for Diffusion Language Models

   Yizhou Liang, Ze Wang, Hao Chen, Ximeng Sun, Jialian Wu, Xiaodong Yu, Jiang Liu, Emad Barsoum, Zicheng Liu

   preprint 2026

3. Instella: Fully Open Language Models with Stellar Performance

   Jiang Liu, Jialian Wu, Xiaodong Yu, et al. (incl. Ze Wang)

   preprint 2025 code

4. SoftVQ-VAE: Efficient 1-Dimensional Continuous Tokenizer

   Hao Chen, Ze Wang, Xiang Li, Ximeng Sun, Fangyi Chen, Jiang Liu, Jindong Wang, Bhiksha Raj, Zicheng Liu, Emad Barsoum

   CVPR 2025

5. Masked Autoencoders Are Effective Tokenizers for Diffusion Models

   Hao Chen, Yujin Han, Fangyi Chen, Xiang Li, Yidong Wang, Jindong Wang, Ze Wang, Zicheng Liu, Difan Zou, Bhiksha Raj

   ICML 2025

6. Binary Latent Diffusion

   Ze Wang, Jiang Wang, Zicheng Liu, Qiang Qiu

   CVPR 2023 code

7. Energy-Inspired Self-Supervised Pretraining for Vision Models

   Ze Wang, Jiang Wang, Zicheng Liu, Qiang Qiu

   ICLR 2023 Spotlight

8. Few-Shot Fast-Adaptive Anomaly Detection

   Ze Wang, Yipin Zhou, Rui Wang, Tsung-Yu Lin, Ashish Shah, Ser-Nam Lim

   NeurIPS 2022

9. Continual Learning with Filter Atom Swapping

   Zichen Miao, Ze Wang, Wei Chen, Qiang Qiu

   ICLR 2022 Spotlight

10. Image Generation Using Continuous Filter Atoms

    Ze Wang, Seunghyun Hwang, Zichen Miao, Qiang Qiu

    NeurIPS 2021 Spotlight

11. Adaptive Convolutions with Per-Pixel Dynamic Filter Atom

    Ze Wang, Zichen Miao, Jun Hu, Qiang Qiu

    ICCV 2021 code

12. Range Adaptation for 3D Object Detection in LiDAR

    Ze Wang, Sihao Ding, Ying Li, Minming Zhao, Sohini Roychowdhury, Andreas Wallin, Guillermo Sapiro, Qiang Qiu

    ICCV-W 2019 Best Paper Award

Full list on Google Scholar.