Skip to content
@THUSI-Lab

Tsinghua Spatial Intelligence Lab (THUSI Lab)

THUSI Lab, part of the College of AI at Tsinghua University, is led by Prof. Yiming Li. Our mission is to scale AI across space 🌍, time ⏳, and embodiments πŸ€–!

Pinned Loading

  1. Hstar Hstar Public

    [CVPR 2026] Thinking in 360Β°: Humanoid Visual Search in the Wild

    Python 150 3

  2. GameVerse GameVerse Public

    [ICML 2026] GameVerse: Can Vision-Language Models Learn from Video-based Reflection?

    Python 55

  3. PrismWAM PrismWAM Public

    Official implementation of PrismWAM: Decoupling State Representations in World Action Models

    Python 62 1

  4. GameScaling GameScaling Public

    Scaling in Games: Continued Pre-Training for Embodied Agents in Diverse Virtual Worlds. Weights on Hugging Face.

    Python 505 29

  5. RoboDoctor RoboDoctor Public

    The official code implementation for RoboDoctor: Training-Free Recursive Self-Improvement for Robot Policies

    Python 19

  6. VSI-Super-Wild VSI-Super-Wild Public

    [ECCV 2026] Towards Spatial Supersensing in the Wild

    Python 22

Repositories

Showing 10 of 10 repositories
  • Awesome-LFMs-Play-Games Public

    Towards Generalist Game Agents: A Survey of Foundation Models in Virtual Worlds

    THUSI-Lab/Awesome-LFMs-Play-Games's past year of commit activity
    Python 53 2 0 0 Updated Nov 9, 2026
  • GameScaling Public

    Scaling in Games: Continued Pre-Training for Embodied Agents in Diverse Virtual Worlds. Weights on Hugging Face.

    THUSI-Lab/GameScaling's past year of commit activity
    Python 505 Apache-2.0 29 2 0 Updated Oct 29, 2026
  • RoboDoctor Public

    The official code implementation for RoboDoctor: Training-Free Recursive Self-Improvement for Robot Policies

    THUSI-Lab/RoboDoctor's past year of commit activity
    Python 19 Apache-2.0 0 0 0 Updated Oct 23, 2026
  • PrismWAM Public

    Official implementation of PrismWAM: Decoupling State Representations in World Action Models

    THUSI-Lab/PrismWAM's past year of commit activity
    Python 62 1 1 0 Updated Oct 23, 2026
  • Causal-Reasoner Public

    Code release for Planning Takes More Than Token Prediction: Causal Plan for Benchmarking and Building Physically Grounded Embodied Reasoners

    THUSI-Lab/Causal-Reasoner's past year of commit activity
    Python 7 0 0 0 Updated Oct 20, 2026
  • VideoArena Public

    Video-Arena: A Scalable Agentic Video Benchmark for Spatiotemporal Capability. Videos and GT on Hugging Face.

    THUSI-Lab/VideoArena's past year of commit activity
    Python 9 Apache-2.0 0 0 0 Updated Oct 14, 2026
  • GameVerse Public

    [ICML 2026] GameVerse: Can Vision-Language Models Learn from Video-based Reflection?

    THUSI-Lab/GameVerse's past year of commit activity
    Python 55 MIT 0 0 0 Updated Oct 7, 2026
  • DAVIS Public

    Official project website for DAVIS, a depth-only active-vision framework for humanoid soccer skills (arXiv:2609.28175).

    THUSI-Lab/DAVIS's past year of commit activity
    HTML 0 0 0 0 Updated Sep 26, 2026
  • Hstar Public

    [CVPR 2026] Thinking in 360Β°: Humanoid Visual Search in the Wild

    Python 150 Apache-2.0 3 0 0 Updated Sep 21, 2026
  • VSI-Super-Wild Public

    [ECCV 2026] Towards Spatial Supersensing in the Wild

    THUSI-Lab/VSI-Super-Wild's past year of commit activity
    Python 22 MIT 0 0 0 Updated Jul 16, 2026