Shuailei Ma
Final-year Ph.D. Student | Northeastern University, China
Shuailei Ma ηš„δΈͺ人照片
I am a final-year Ph.D. student at Northeastern University, China. My research interests mainly lie in pre-training visual generative foundation models. I am also involved in pre-training world models, Vision-Language-Action models (VLA), and Video-Action models (VA).

I developed a large-scale pre-training framework for video diffusion models that achieves over 50% Model FLOPs Utilization (MFU) on a thousand-GPU cluster. Building on this framework, we developed LingBot-Video, the first open-source state-of-the-art video generation model based on a sparse Mixture-of-Experts (MoE) architecture. We also used this framework to explore the scaling behavior of video diffusion models, scaling up to 120B total parameters with 11B active parameters under our available compute budget.

πŸ”¬ I am currently in the final year of my Ph.D. and working as a research intern at Robbyant, an embodied AI company under Ant Group, where I work with Kecheng Zheng.

πŸŽ“ Education

πŸ“° News

πŸ—‚οΈ Selected Projects

β™  Generation & World Models 🎬 Scaling Mixture-of-Experts Video Pretraining for Embodied Intelligence 🌍 Advancing Open-source World Models ✨ Learning Visual Generative Priors without Text
🎬 Scaling Mixture-of-Experts Video Pretraining for Embodied Intelligence
Shuailei Ma, Jiaqi Liao, Xinyang Wang, Jingjing Wang, Chaoran Feng, Zijing Hu, Chong Bao, Zichen Xi, Yuqi Gan, Weisen Wang, Yanhong Zeng, Qin Zhao, Zifan Shi, Wei Wu, Hao Ouyang, Qiuyu Wang, Shangzhan Zhang, Jiahao Shao, Yipengjing Sun, Liangxiao Hu, Lunke Pan, Nan Xue, Kecheng Zheng, Yinghao Xu, Xing Zhu, Yujun Shen, Ka Leong Cheng
A DiT-based video pre-training paradigm with sparse MoE, robot-oriented data, and multi-dimensional rewards for embodied intelligence.
Technical ReportPaperCodeProject Page
🌍 Advancing Open-source World Models
Robbyant Team, Zelin Gao, Qiuyu Wang, Yanhong Zeng, Jiapeng Zhu, Ka Leong Cheng, Yixuan Li, Hanlin Wang, Yinghao Xu, Shuailei Ma, Yihang Chen, Jie Liu, Yansong Cheng, Yao Yao, Jiayi Zhu, Yihao Meng, Kecheng Zheng, Qingyan Bai, Jingye Chen, Zehong Shen, Yue Yu, Xing Zhu, Yujun Shen, Hao Ouyang
An open-source world simulator with high-fidelity dynamics, minute-long consistency, and real-time interaction.
Technical ReportPaperCodeProject Page
β™  Vision-Language-Action & Video-Action Models πŸ€– A Pragmatic VLA Foundation Model 🦾 From Foundation to Application: Improving VLA Models in Practice 🎞️ Native Video-Action Pretraining for Generalizable Robot Control
πŸ€– A Pragmatic VLA Foundation Model
Wei Wu, Fan Lu, Yunnan Wang, Shuai Yang, Shi Liu, Fangjing Wang, Qian Zhu, He Sun, Yong Wang, Shuailei Ma, Yiyu Ren, Kejia Zhang, Hui Yu, Jingmei Zhao, Shuai Zhou, Zhenqi Qiu, Houlong Xiong, Ziyu Wang, Zechen Wang, Ran Cheng, Yong-Lu Li, Yongtao Huang, Xing Zhu, Yujun Shen, Kecheng Zheng
A pragmatic VLA foundation model pre-trained on around 20,000 hours of real-world data from nine popular dual-arm robot configurations.
Technical ReportPaperCodeProject Page
🦾 From Foundation to Application: Improving VLA Models in Practice
Wei Wu, Fangjing Wang, Fan Lu, He Sun, Shi Liu, Yunnan Wang, Yibin Yan, Yong Wang, Shuailei Ma, Xinyang Wang, Yibin Liu, Shuai Yang, Tianxiang Zhou, Kejia Zhang, Lei Zhou, Cheng Su, Nan Xue, Bin Tan, Han Zhang, Youchao Zhang, Fei Liao, Xing Zhu, Yujun Shen, Kecheng Zheng
LingBot-VLA 2.0 improves task and embodiment generalization, expands whole-body action spaces, and introduces predictive dynamics modeling.
Technical ReportPaperCodeProject Page
🎞️ Native Video-Action Pretraining for Generalizable Robot Control
Qihang Zhang, Lin Li, Luyao Zhang, Shuai Yang, Yiming Luo, Shuaiting Li, Ruilin Wang, Junke Wang, Jiahao Shao, Gangwei Xu, Jiaming Zhou, Yishu Shen, Yudong Jin, Fangyi Xu, Shuailei Ma, Jiaqi Liao, Guanxing Lu, Zifan Shi, Yongkun Wen, Yujie Zhao, Weixuan Tang, Xinyang Wang, Chaojian Li, Jiapeng Zhu, Ka Leong Cheng, Nan Xue, Xing Zhu, Yujun Shen, Yinghao Xu
A video-action foundation model natively pre-trained for embodiment, featuring a semantic visual-action tokenizer, causal pre-training, sparse MoE, and real-time closed-loop control.
Technical ReportPaperCodeProject Page
β™  Multi-Modal Large Language Models 🐬 Aligned Better, Listen Better for Audio-Visual Large Language Models 🧩 CoReS: Orchestrating the Dance of Reasoning and Segmentation 🏠 LSceneLLM: Enhancing Large 3D Scene Understanding Using Adaptive Visual Preferences
β™  Object Detection βš“ FGAHOI: Fine-Grained Anchors for Human-Object Interaction Detection πŸ”­ CAT: LoCalization and IdentificAtion Cascade Detection Transformer for Open-World Object Detection 🧠 SKDF: A Simple Knowledge Distillation Framework for Distilling Open-Vocabulary Knowledge to Open-World Object Detector

πŸ§‘β€πŸ’» Professional Activities