🇨🇭phd student @EPFL_en | working on embodied ai and computational neuroscience | previous intern at deepmind, bytedance ai lab | open science
Switzerland
Joined July 2013
- Tweets395
- Following335
- Followers540
- Likes1.7K
Chengkun Li retweeted
What if a simulated human moved with real muscles instead of motors? 🦾
MuscleMimic: one RL policy controls 416 muscles for walking, running, dancing, lifting. Fully open source!
youtube.com/watch?v=dvoZbccT…
#ReinforcementLearning #EmbodiedAI #Robotics #AI
Chengkun Li retweeted
This is a super exciting time for science, and EPFL is growing.
Multiple faculty openings in areas I’m especially excited about:
• Computer science — broad search
• Security & privacy
• Mathematics of learning
• Ethics of technology — including alignment & algorithmic bias
Please consider applying, and help us spread the word! 👇
Chengkun Li retweeted
🏌️♂️ Controlling 354 muscles of human body with a musclemimic generalist; zero-shot on chained movements
To train yourself -> github.com/amathislab/muscle…
Chengkun Li retweeted
The Bitter Lesson for harness design and formal math!
Instead of hand-designing how an AI should prove (decompose, plan, assign agents), what if we just give a frontier model good Lean feedback and let it keep working?
We built h1.
TL;DR:
→ SOTA on formal theorem proving benchmarks
→ can solve open conjectures
→ matches/beats hand-designed workflows
→ one long attempt can beat many retries
[1/7]
Chengkun Li retweeted
PhD Advising vs Vibe Coding #1:
One of the key skills a PhD advisor learns is context switching. Students come into your office every 30-60 minutes and you need to remember their project, where you left off, and what they were supposed to do next. It's a learned skill and exactly what you need to manage AI agents.
Chengkun Li retweeted
⛰️TERRA is a very interesting and impressive work led by Merkourios Simos @amathislab
It substantially expands the terrain diversity we can train and evaluate MuscleMimic controllers on, going beyond flat-ground motion to diverse, non-flat environments reconstructed directly from motion data.
Huge thanks to folks who's work made this possible, @CaggianoVitt and the @MyoSuite team, @CherylWang39256, Bianca Ziliotto, @TrackingPlumes and MuscleMimic team, and the amazing teams at @GoogleDeepMind and @nvidia behind mujoco warp.
Exciting to see how quickly the field is progressing, and especially what this opens up for biomechanics and neuroscience. Please checkout our paper at: arxiv.org/abs/2609.38653
Chengkun Li retweeted
and of course, choreography dance with muscles :)
🏌️♂️ Controlling 354 muscles of human body with a musclemimic generalist; zero-shot on chained movements
To train yourself -> github.com/amathislab/muscle…
Chengkun Li retweeted
🌐Does modeling dynamics provide a better path to generalization than learning model-free policies alone?
🐕In QWM @corl_conf, a morphology-conditioned world model learns shared locomotion dynamics and enables zero-shot transfer to unseen morphologies.
🔗modanesh.github.io/papers/qw…
The code is released now on github!
Try it now -> github.com/amathislab/terra
Learn more at: cnai.epfl.ch/terra
⛰️TERRA is a very interesting and impressive work led by Merkourios Simos @amathislab
It substantially expands the terrain diversity we can train and evaluate MuscleMimic controllers on, going beyond flat-ground motion to diverse, non-flat environments reconstructed directly from motion data.
Huge thanks to folks who's work made this possible, @CaggianoVitt and the @MyoSuite team, @CherylWang39256, Bianca Ziliotto, @TrackingPlumes and MuscleMimic team, and the amazing teams at @GoogleDeepMind and @nvidia behind mujoco warp.
Exciting to see how quickly the field is progressing, and especially what this opens up for biomechanics and neuroscience. Please checkout our paper at: arxiv.org/abs/2609.38653
and of course, choreography dance with muscles :)
🏌️♂️ Controlling 354 muscles of human body with a musclemimic generalist; zero-shot on chained movements
To train yourself -> github.com/amathislab/muscle…
Chengkun Li retweeted
We’re releasing a broad range of new mathematical results produced by an internal frontier model.
We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results.
github.com/openai/math
Chengkun Li retweeted
Can a humanoid learn robust whole-body loco-manipulation directly from a human? No robot, no teleop, no retargeting.
Meet Workhorse 🐴: our G1 topples and climbs a suitcase, catches a thrown box, and sorts boxes with hands and a kick. Fully autonomous.
More videos at hybridrobotics.github.io/wor…
Chengkun Li retweeted
Tu Youyou became the first mainland Chinese scientist to be awarded a #NobelPrize in a scientific field - for discovering artemisinin, a malaria cure that’s saved millions. Today we reveal the 2026 medicine laureate. Stay tuned.
ALT A portrait of Tu Youyou in front of a dark grey background. The light is on the left side of her face leaving the right side slightly in shadow. She is looking directly into the camera with a small smile.
Chengkun Li retweeted
We taught a skeleton 💀 to dance. Not by animating it, but letting it learn the full 354 simulated muscle control. 💪🏼
—> teach them yourself at: github.com/amathislab/muscle…
Chengkun Li retweeted
World clocks have been boring for way too long.
Stop doing timezone math. Just drag the sun. ☀️ → 🌙
Check it out: apps.apple.com/us/app/hourli…
Chengkun Li retweeted
in the age of the machines history will unfold in minutes, then seconds, with titanic ideological battles being fought, whole lineages being replaced in a blink. it’s hard to know how to feel about it; the heart sings as an observer and quails as a participant
🏌️♂️ Controlling 354 muscles of human body with a musclemimic generalist; zero-shot on chained movements
To train yourself -> github.com/amathislab/muscle…