Robotic Research @ NVIDIA GEAR PhD @ NTU

Sicong Leng冷思聪

Towards generalist intelligence for the physical world.

I'm a fourth-year PhD student at NTU, advised by Prof. Shijian Lu, and work on robotics at NVIDIA GEAR.

I study how machines can learn from diverse experience, reason beyond what they have seen, and turn understanding into useful action.

Sicong Leng
Multimodal & embodied intelligence

01 / Research

Selected work

Scaling how machines learn, reason, and act.

Robotics data engines2026

Flora

Making robot learning scale with experience.

Turning human and robot video into action captions and rewards, so rich supervision can grow with the data.

First authorTechnical report 2026
All authors

Sicong Leng, Yunze Man, Zhiqi Li, Jing Wang, Yuhe Zhang, Ryan Punamiya, Siddharth Patki, Kaiyuan Zheng, Zi-ang Cao, Zhiding Yu, Sifei Liu, Danfei Xu, Ben Levin, Jan Kautz, Yuke Zhu, Johan Bjorck, Linxi "Jim" Fan

Paper Project Models

Embodied foundation models2026

RynnBrain 1.1

Towards robots that generalize beyond a single task.

Connecting perception, spatial reasoning, and action in a shared foundation for learning across tasks and embodiments.

Core contributorarXiv 2026
All authors

Kehan Li, Bohan Hou, Minghao Zhu, Tianyi Zhang, Zesen Cheng, Zhikai Wang, Sicong Leng, Xin Li, Xiao Lin, Biying Yao, Minghua Zeng, Jiangpin Liu, Ronghao Dang, Jiayan Guo, Siteng Huang, Haoyu Zhao, Heng Ping, Yaxi Zhao, Tong Zhao, Kexiang Wang, Tong Lu, Shengke Xue, Jiahao Tang, Yulei Wang, Zejing Wang, Jianwei Gao, Shijian Lu, Chengju Liu, Jianfei Yang, Mingxiu Chen, Deli Zhao

Paper Project Code Models

More selected work

Publication archive

For the full publication list, visit Google Scholar ↗.

02 / Updates

Recently

New work and open releases.

  1. Introducing Flora: open models for timestamped action captions and reward signals, powering the robotics data engine for GR00T.

  2. RynnBrain 1.1 is open source, with stronger embodied understanding, contact-point prediction, and native 3D grounding.

  3. Vesta brings localization, navigation, reasoning, and planning into one embodied generalist.

More updates

  1. New survey out: World Model for Robot Learning: A Comprehensive Survey — a policy-centric look at world models for robot learning.

  2. New survey out: Visual Generation in the New Era: An Evolution from Atomic Mapping to Agentic World Modeling — a roadmap from atomic mapping to agentic world modeling.

  3. MMR1 and LongVT accepted by CVPR 2026.

  4. RynnBrain is released — open embodied foundation models for robotics.

  5. RynnVLA-001 accepted by ICRA 2026.

  6. MMR1 is out — variance-aware sampling for multimodal reasoning, with open-sourced code, data, and model.

  7. CMM and Better MoE accepted by NeurIPS 2025 — Better MoE as oral.

  8. RynnVLA-001 is released — a vision-language-action model pretrained on human demonstrations for robot manipulation.

  9. Inf-CLIP received the CVPR 2025 Highlight.

  10. MMR1 has been released — initial code drop for our multimodal reasoning research.

  11. Inf-CLIP and AGLA accepted by CVPR 2025.

  12. VideoLLaMA 3 has been released.

  13. Inf-CLIP has been released! Project.

  14. CMM has been released! Project.

  15. 1 paper accepted by NeurIPS 2024!

  16. VideoLLaMA 2 has been released! Paper & code.

  17. VCD received the CVPR 2024 Highlight.

  18. 3 papers accepted by CVPR 2024!

  19. VCD has been released! Paper & code.

  20. Presented our work at the Nvidia Internal Technical Sharing.

  21. Presented our work at the AAAI 2023 Summer Symposium Series.

  22. Tell2Design received the Area Chair Award and Best Paper Nomination at ACL 2023!

  23. Tell2Design accepted by ACL 2023 as a long oral paper.

03 / Background

Along the way

Education

Nanyang Technological
University

PhD · Fourth year

Advised by Prof. Shijian Lu.

May 2026 – Present

NVIDIA GEAR

Robotic Research

Building GR00T through robotics data engines and embodied reasoning.

Aug 2023 – May 2026

Alibaba DAMO Academy

Embodied Intern

Multimodal models and robot learning: VideoLLaMA, MMR1, RynnVLA, and RynnBrain. Alibaba Talent Program.

With Xin Li , Deli Zhao

Sep 2021 – Aug 2023

StatNLP Lab, SUTD

Research Officer

NLP and multimodal learning, including language-guided floor plan generation with Tell2Design.

With Prof. Lu Wei

“We were born too late to explore the earth and too early to explore the stars. We were born just in time to solve robotics. I'm on a mission to make Physical AI a reality, one motor at a time.”

— Jim Fan

Let's connect.

Away from research: badminton, singing, and games.