Hongyi Zhang

Physical AGI @ PokeBot

[Hongyi Zhang @Hongyi_Zhang]

Would GPT-6-Astra release be the most important event of 2026 for physical AI?

[thijs @cdngdev]

i gave astra a robot, a paint brush, and a camera then asked it to paint the golden gate bridge in real life!

it figured out how to control the robot, and progressively got better throughout its attempts. the timelapse is sick

[Hongyi Zhang @Hongyi_Zhang]

good insights. but latency still bites for reactive policies?

[Michael Zeng @michzeng_]

Action chunking — especially executing long action sequences open-loop — is widely used in imitation learning for robotic manipulation. Why is it so effective and do we really need it? We find a key reason:

Long open-loop execution helps short-context policies imitate non-Markovian experts.

With this insight, we show how to move beyond open-loop execution: extending policy context restores reactivity while achieving even higher task performance.

🧵(1/5)

[Suvrit Sra @optiML]

Gradient descent extends easily to Riemannian manifolds. Does Nesterov's accelerated gradient method generalize to Riemannian manifolds? The answer turns out to be rather intricate. First set of answers in our colt18 paper: https://t.co/bXZNnA19Ds

[Hongyi Zhang @Hongyi_Zhang]

cong ;) on holidays? @xlnxyx

[Hongyi Zhang @Hongyi_Zhang]

@eyelid_s same with me... I'm writing SoP one word at a time...

[Hongyi Zhang @Hongyi_Zhang]

If you have something fantastic, try to understand it; if you cannot understand it, try not to use it.

[Hongyi Zhang @Hongyi_Zhang]

@eyelid_s ...@@

[Dan Roy @roydanroy]

We got a ton of high quality submissions to our #NIPS2012 Workshop on Probabilistic Programming. Going to be awesome. http://t.co/y8NwOvsT

[Hongyi Zhang @Hongyi_Zhang]

maybe the machine learning community is flooded by overly optimistic and enthusiastic youth not even knowing how far they can possibly reach

[Hongyi Zhang @Hongyi_Zhang]

Hello, World!