Adam Wei

@adamwei_

Intern @physical_int 🏖️, Robotics PhD student @MIT_CSAIL with Prof. @RussTedrake

Joined April 2025

315 Following

705 Followers

Replies Reposts Media

Pinned
Adam Wei
@adamwei_
Jun 18
🤖 We introduce Ambient Diffusion Policy, a simple and principled method for training policies with suboptimal data in robotics. Suboptimal data is everywhere in robotics… ❌ Data filtering is wasteful ❌ Co-training learns both good and bad features ✅ Ambient Diffusion Policy
17
Adam Wei
@adamwei_
Sep 4
🎉 Ambient Diffusion Policy was accepted to CoRL 2026! See y'all in Austin 🫡
Adam Wei
@adamwei_
Jun 18
🤖 We introduce Ambient Diffusion Policy, a simple and principled method for training policies with suboptimal data in robotics. Suboptimal data is everywhere in robotics… ❌ Data filtering is wasteful ❌ Co-training learns both good and bad features ✅ Ambient Diffusion Policy
1
Adam Wei
@adamwei_
Aug 21
Great work from my labmate @michzeng_ on action-chunking! Def worth a read, especially the investigation into the interplay between context length and action horizon
Michael Zeng
@michzeng_
Aug 20
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
1
Adam Wei
@adamwei_
Jul 24
Ambient Diffusion Policy won 2/2 Best Paper Awards at #RSS2026 workshops! 🏆🏆 • Best Paper — Data-Centric Robotics • Most Useful Practical Information — It’s the Demos I'm super excited to see how the community builds on this work. Huge thanks to the organizers for having
Adam Wei
@adamwei_
Jun 18
00:41
🤖 We introduce Ambient Diffusion Policy, a simple and principled method for training policies with suboptimal data in robotics. Suboptimal data is everywhere in robotics… ❌ Data filtering is wasteful ❌ Co-training learns both good and bad features ✅ Ambient Diffusion Policy
2
Adam Wei
@adamwei_
Jul 22
Checkout this detailed investigation into context length and a new variable history curriculum training method!
Abhinav Agarwal
@_abhinav_ag
Jul 22
Roboticists have been working hard to add memory to “goldfish” imitation learning policies. But do we understand how context length impacts state-of-the-art imitation learning methods? We provide an in-depth analysis of long-context Diffusion Policies to study this 🧵(1/7)

Adam Wei

Adam Wei’s posts