Assistant Professor at Tsinghua University

• Pinned

  

  Tennis is an extremely challenging sport — even for humans. Yet together with

  @GalbotRobotics

  , we manage to teach humanoid robots to play it and sustain long rallies with human players. More importantly, it precisely returns incoming balls while maintaining highly human-like

  

  00:00

• 

  World’s first humanoid robot tennis open exhibition match — livestreamed globally with no backup footage. Galbot robots shared the court with a world tennis champion in a high-level showcase. After a week at the National Speed Skating Oval, we delivered an outstanding

  

  From AlphaGo to AstraTennis. In 2016, AlphaGo showed AI could master the digital world. A decade later, AI enters the physical world. Today, Galbot robots achieved the world’s first live autonomous humanoid robot tennis match, competing against human athletes at the World

  

  00:00

• 

  LDA is built on a simple belief: embodied intelligence should not waste data. Instead of relying only on scarce high-quality expert demonstrations, we show that robots can learn from all kinds of embodied data—and scale from it, just like GPT did for language. Congrats to all

  

  Introducing LDA, a latent world action foundation model that, for the first time, unifies the utilization of heterogeneous embodied data across simulation and reality, humans and robots, and varying levels of action quality and annotation. By breaking long-standing data silos in

  

  00:00

• 

  ⏳ Tick Tock! The 3DV 2026 Nectar Track deadline is almost here! You have until Feb 15 to submit your best work. We want to see what you've been working on! Choose your track: 1️⃣ ✨ Spotlight Track (Oral + Poster) 2️⃣ 🧭 Exploration Edge Track (Poster) Get those submissions in!

  

  The Nectar Track deadline is fast approaching! (Feb. 15) ⏳ Share your amazing work in either the ✨ Spotlight Track (Oral + Poster) or the 🧭 Exploration Edge Track (Poster)! See the link below ⬇️

  

• 

  Thrilled to share one of my favorite works this year: DexNDM! We bridge the Sim2Real gap for dexterous in-hand rotation, achieving a true "0-to-1" advancement. The key? DexNDM learns from biased, real-world data without needing any successful demonstrations. Now a

  

  00:00