Gao Jiawei
46 posts
Gao Jiawei
@WinstonGu_
Final Year Undergrad at @Tsinghua_Uni; Currently @Harvard; Previously @CMU_Robotics; Robot Learning & Embodied Agents. Applying for PhD at 2026 Fall!
Katılım Nisan 2022
712 Takip Edilen538 Takipçiler
Check out our new work on an interesting action tokenizer design! ordered-action-tokenization.github.io
Chaoqi's insightful post is definitely worth reading.
Chaoqi Liu@liu730chaoqi
English
@jbhuang0604 From the perspective of an PhD applicant, it is hard to apply for a phd position these years given that many faculties are actually on-leave.
English
We saw a significant drop in the number of CS PhD applicants this year 🤯
The decline is even sharper for computer vision! 😱
Severin Hacker@severinhacker
Should you get a PhD in CS/AI? Both seem to be true: 1. 95% of PhDs would have done equally well career-wise without one. 2. 95% of real AI breakthroughs (ImageNet, Transformers) came from PhDs.
English
This matches my intuition! Check out this work on scientifically and systematically understanding Flow Policies vs. Regression Policies in Robotics!
Chaoyi Pan@ChaoyiPan
Generative models (diffusion/flow) are taking over robotics 🤖. But do we really need to model the full action distribution to control a robot? We suspected the success of Generative Control Policies (GCPs) might be "Much Ado About Noising." We rigorously tested the myths. 🧵👇
English
Just reconstructed a 3D model of Porcelain Tower of Nanjing(南京大報恩寺琉璃塔) using SAM 3D. The tower was burned down in 1856, which means it would in no way exist in the training data.
The result is CRAZY.
English
@JieWang_ZJUI @physical_int @sundayrobotics @hausman_k This is just espresso, by the way. I’m wondering if the current policies can froth the milk.
English
Making latte is much harder for robot than it looks like!
Congrats for @physical_int and @sundayrobotics achieving this goal.
However, based on my experience, smaller machine is harder for robot to play with, @hausman_k could we try pi*06 on Breville too?
Karol Hausman@hausman_k
This is one of my favorite videos -- makes you appreciate how difficult it is to make a latte (wait for the wrist cam footage towards the end)
English
@ChaoyiPan Impressive work! Hopefully one day we can retarget robot-spider motions using SPIDER!
English
🕸️ Introducing SPIDER — Scalable Physics-Informed Dexterous Retargeting!
A dynamically feasible, cross-embodiment retargeting framework for BOTH humanoids 🤖 and dexterous hands ✋.
From human motion → sim → real robots, at scale.
🔗 Website: jc-bao.github.io/spider-project/
🧵 1/n
English
🤫 Something's been brewing in stealth. Our SDK team's side project, codenamed W&B LEET, is being unleashed.
We are releasing a full Terminal UI (TUI) for live, interactive W&B monitoring right in your terminal.
No browser, no internet, no problem.
English
Excited to introduce COLA: Learning Human–Humanoid Coordination for Collaborative Object Carrying 🤝🤖
COLA makes humanoids truly helpful in human collaboration — capable of carrying objects, pushing carts, or responding to human push commands.
It provides a proprioception-only policy for compliant human–humanoid coordination across diverse movement patterns.
The core idea is simple yet effective:
👉 Fine-tune a collaborative policy from a locomotion policy using a residual teacher.
👉 Train in simulation and distill to a real-world student policy for deployment.
Paper: arxiv.org/abs/2510.14293
Project: yushi-du.github.io/COLA/
English
It would be really exciting if someone could build on this work for humanoid robots! Original code at github.com/Winston-Gu/Coo…
Gao Jiawei@WinstonGu_
Imagine a future where you can ask humanoid robots to clean your room, but some items, like heavy sofas, are too challenging for just one robot to move. Introducing CooHOI, a learning-based framework designed for the cooperative transportation of objects by multiple humanoid robots. 🤖🤼🤖 Our work has been accepted as Spotlight at NeurIPS 2024. Website: gao-jiawei.com/Research/CooHO…
English
RoboDuet has been accepted by IROS for presentation!
See you in Hangzhong this October!
😭Really eager to be there in person
Elgce@BenQingwei
After a long time, my first research project of my life, RoboDuet, has finally been accepted by RAL! This is really inspiring for me. RoboDuet is fully open-sourced, including training code and deployment code. If you're interested in it, just have a try!
English
Learned a lot from this insightful post! Really hope there will be more posts like this in the community.
Seohong Park@seohong_park
Q-learning is not yet scalable seohong.me/blog/q-learnin… I wrote a blog post about my thoughts on scalable RL algorithms. To be clear, I'm still highly optimistic about off-policy RL and Q-learning! I just think we haven't found the right solution yet (the post discusses why).
English
Got my tenure! Very grateful to my students and collaborators.
English
BTW, I passed!!!
Graduated with a PhD after 6 amazing years at CMU’s Robotics Institute!
Huge thanks to my advisor @kkitani, my committee members @shubhtuls @GuanyaShi @xbpeng4 Karen Liu, and everyone who has supported me!!!
Checkout my defense talk: Learning Universal Humanoid Control → youtu.be/tG0mbs_u_ik
#CMU #RI #Robotics #PhD #Graduation #Humanoids
YouTube
Zhengyi “Zen” Luo@zhengyiluo
🎓 Excited to defend my PhD thesis “Learning Universal Humanoid Control” at CMU this Friday! From scalable motion imitators to visual dexterous whole-body policies — it’s been a wild ride 🤖✨ 📅 April 25, 2025 📍 CMU RI & online 🔗 cs.cmu.edu/calendar/18255…
English
Training robots for the open world needs diverse data
But collecting robot demos in the wild is hard!
Presenting DexWild
🙌🏕️ Human data collection system that works in diverse environments, without robots
💪🦾 Human + Robot Cotraining pipeline that unlocks generalization
🧵👇
English
🤖Introducing TWIST: Teleoperated Whole-Body Imitation System.
We develop a humanoid teleoperation system to enable coordinated, versatile, whole-body movements, using a single neural network.
This is our first step toward general-purpose robots.
🌐humanoid-teleop.github.io
English
It has been a wonderful journey with Ani over the years, and I couldn’t have imagined a more brilliant and supportive advisor and friend. Ani shed light on my research when I knew little in robot learning and reminded me to always think beyond the envelope. All the best Ani!
Anirudha Majumdar@Majumdar_Ani
Congratulations Dr. Allen Ren @allenzren! What an incredible honor it's been to have you in our lab over the past 5.5 years, and to learn from you. Very excited to follow your next steps in bringing general-purpose AI into the physical world!
English