1. X
  2. Kaifeng Zhang
Log inSign up
Kaifeng Zhang
289 posts
Image
user avatar
Kaifeng Zhang
@kaiwynd
PhD student at Columbia University
New York, NY
kywind.github.io
Joined May 2015
1,181
Following
2,793
Followers
RepliesRepliesMediaMedia

New to X?

Sign up now to get your own personalized timeline!

Create account

By signing up, you agree to the Terms of Service and Privacy Policy, including Cookie Use.

Terms·Privacy·Cookies·Accessibility·Ads Info·© 2026 X Corp.
Don't miss what's happening
People on X are the first to know.
Log inSign up
  • Pinned
    user avatar
    Kaifeng Zhang
    @kaiwynd
    Feb 5
    Real2Sim-Eval has been accepted to #ICRA2026 ! 🎉 Transform your real setup into GS-based simulation environment, and evaluate robot policies. Website: real2sim-eval.github.io arXiv: arxiv.org/pdf/2511.04665 Code: github.com/kywind/real2si…
    user avatar
    Kaifeng Zhang
    @kaiwynd
    Nov 6, 2025
    🧵 Evaluating robot policies in the real world is slow, expensive, and hard to scale. During my internship at @SceniXai this summer, we had many discussions around the two key questions: how accurate must a simulator be for evaluation to be meaningful, and how do we get there?
    Image
    00:00
    9K
  • user avatar
    Kaifeng Zhang
    @kaiwynd
    Nov 7, 2025
    If you are working on real-to-sim, simulating digital twins, and policy evaluation, you should check out our fully open-sourced code base. Lots of handy tools for building Gaussian Splatting simulators and interacting with it! github.com/kywind/real2si… Will continue to be
    Image
    GitHub - kywind/real2sim-eval: Open-source code of the paper: Real-to-Sim Robot Policy Evaluation...
    From github.com
    16K
  • user avatar
    Kaifeng Zhang
    @kaiwynd
    Jun 19, 2025
    Can we learn a 3D world model that predicts object dynamics directly from videos? Introducing Particle-Grid Neural Dynamics: a learning-based simulator for deformable objects that trains from real-world videos. Website: kywind.github.io/pgnd ArXiv: arxiv.org/abs/2506.15680
    Image
    00:00
    46K
  • user avatar
    Kaifeng Zhang
    @kaiwynd
    Nov 6, 2025
    🧵 Evaluating robot policies in the real world is slow, expensive, and hard to scale. During my internship at @SceniXai this summer, we had many discussions around the two key questions: how accurate must a simulator be for evaluation to be meaningful, and how do we get there?
    Image
    00:00
    62K
  • user avatar
    Kaifeng Zhang
    @kaiwynd
    Apr 11, 2023
    Finally the moment. I will be joining the University of Illinois @illinoiscs as a PhD student this fall! I will be working with Prof. Yunzhu Li @YunzhuLiYZ on exciting topics across computer vision, machine learning and robotics.
    Image
    32K
  • user avatar
    Kaifeng Zhang
    @kaiwynd
    Aug 4, 2025
    Excited to see that particle-based dynamics model learning + MPC gets applied in cloth manipulation with multifingered hands! The idea of predicting object motions by representing them as dense particles shares the same insight as our recent paper, PGND
    user avatar
    Harry Li
    @Haoyang1i
    Aug 2, 2025
    Modeling the complex 3D dynamics of deformable objects is a critical challenge in robotics. How can we achieve accurate long-horizon prediction for tasks like cloth manipulation 👕? Thrilled to announce that our work, "UniClothDiff," has been accepted to #CoRL2025! We leverage a
    Image
    00:00
    9.1K
  • user avatar
    Kaifeng Zhang
    @kaiwynd
    Jun 27, 2025
    In his RSS keynote talk, Trevor Darrell predicted that 4D particle will be the most effective representation for visual pretraining. Exciting to see this vision align with the core design of our work PGND: learning (4D) particle dynamics from real-world RGB-D videos. 👀
    Image
    Image
    00:34
    user avatar
    Kaifeng Zhang
    @kaiwynd
    Jun 19, 2025
    Can we learn a 3D world model that predicts object dynamics directly from videos? Introducing Particle-Grid Neural Dynamics: a learning-based simulator for deformable objects that trains from real-world videos. Website: kywind.github.io/pgnd ArXiv: arxiv.org/abs/2506.15680
    8.4K
  • user avatar
    Kaifeng Zhang
    @kaiwynd
    Oct 28, 2025
    Hard agree — with a good 3D Gaussian Splatting model, sim and real can look almost identical! Check out this amazing paper! We also have a recent work that builds highly realistic simulations — both visually and dynamically (an often overlooked aspect!). Release coming soon —
    user avatar
    Xiaolong Wang
    @xiaolonw
    Oct 28, 2025
    The Real2Sim is so good that it takes some time to tell what is real and what is sim.
    18K
  • user avatar
    Kaifeng Zhang
    @kaiwynd
    Jul 11, 2024
    (1/8) Introducing AdaptiGraph, our new paper accepted by #RSS2024 We show that a GNN dynamics model can model deformable objects with varying physical properties. Website: robopil.github.io/adaptigraph/ ArXiv: arxiv.org/abs/2407.07889 Work done w/ @BaoyuLi6, Kris Hauser, @YunzhuLiYZ
    Image
    00:00
    41K
  • user avatar
    Kaifeng Zhang
    @kaiwynd
    Jun 30, 2023
    Officially graduated (under 40 ℃ weather)
    Image
    Image
    9.3K
  • user avatar
    Kaifeng Zhang
    @kaiwynd
    Oct 9, 2025
    The pursuit of a unified benchmark for robot manipulation has been long-standing — and with ManipulationNet, it’s finally taking a leap forward! manipulation-net.org ManipulationNet is a fully real-world, verifiable, and scalable benchmark for robotic manipulation. Anyone
    Image
    Image
    user avatar
    Yiting Chen
    @YitingChen07
    Oct 8, 2025
    “You can’t make progress until you are able to measure it. Robotics still doesn’t have such a rallying call. No one agrees on anything.” I 💯 agree with the recent post from @DrJimFan. To break this impasse, we are excited to announce ManipulationNet (manipulation-net.org), a
    9K
  • user avatar
    Kaifeng Zhang
    @kaiwynd
    Jul 13, 2024
    I will be at #RSS2024 next week to present this paper! Looking forward to my first in-person conference experience, and hope to meet lots of new and old friends :D
    user avatar
    Kaifeng Zhang
    @kaiwynd
    Jul 11, 2024
    (1/8) Introducing AdaptiGraph, our new paper accepted by #RSS2024 We show that a GNN dynamics model can model deformable objects with varying physical properties. Website: robopil.github.io/adaptigraph/ ArXiv: arxiv.org/abs/2407.07889 Work done w/ @BaoyuLi6, Kris Hauser, @YunzhuLiYZ
    Image
    00:00
    2.9K
  • user avatar
    Kaifeng Zhang
    @kaiwynd
    Nov 4, 2024
    Heading to CoRL tomorrow to present this work ✈️
    user avatar
    Kaifeng Zhang
    @kaiwynd
    Oct 25, 2024
    Beyond just reconstruction, we model the physics/dynamics of deformable objects with 3D Gaussians.✨ Excited to share our latest work at #Corl2024. Check it out! gs-dynamics.github.io (1/5)
    1.9K
  • user avatar
    Kaifeng Zhang
    @kaiwynd
    Oct 25, 2024
    Beyond just reconstruction, we model the physics/dynamics of deformable objects with 3D Gaussians.✨ Excited to share our latest work at #Corl2024. Check it out! gs-dynamics.github.io (1/5)
    user avatar
    Mingtong Zhang
    @alexzhang_robo
    Oct 25, 2024
    🚀 Watch robots predict the future! 🤖🔮 Our latest research in action-conditioned video prediction utilizes 3D Gaussian Splatting with graph-based neural dynamics, providing robots the power to predict how objects behave with interactions! 🎥 This paper has been accepted to CoRL
    Image
    00:00
    4K
Advertisement
Advertisement