Pinned Tweet
Introducing ABC: open data, training, and infrastructure for robotics.
We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques.
@arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcall
We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques.
@arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcall
34
113
734
372,109
Ritvik Singh retweeted
Excited to introduce QF3: Fast Flow RL with Filtered Q-Gradients!
Train humanoids from scratch, fine-tune VLAs, even fine-tune image models... all with one simple off-policy update ✨
Train humanoids from scratch, fine-tune VLAs, even fine-tune image models... all with one simple off-policy update ✨
13
59
424
39,372
Ritvik Singh retweeted
Happy to announce the full ABC release! We’re also excited that ABC was accepted to CoRL 2026!
Check out our website for code, 400+ hours of sim data on 24 tasks, and 5,850 labeled policy-evaluation episodes.
@arthurallshire @Cinnabar233 @ritvik_singh9 @redstone_hong @davidrmcall
Check out our website for code, 400+ hours of sim data on 24 tasks, and 5,850 labeled policy-evaluation episodes.
@arthurallshire @Cinnabar233 @ritvik_singh9 @redstone_hong @davidrmcall
6
33
259
77,425
Ritvik Singh retweeted
Excited to share SPD: simulation pre-training for dexterity. We pre-trained a policy in simulation and fine-tuned with less than 2 hours of real data (with @sarthakkamat)
17
71
539
139,735
Ritvik Singh retweeted
ABC Box is available now at i2rt.com
Introducing ABC: open data, training, and infrastructure for robotics.
We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques.
@arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcall
7
9
117
25,425
Ritvik Singh retweeted
🤖 How can we teach dexterous robots to perform precise, contact-rich assembly?
Introducing Play2Perfect: first learn to play with objects, then perfect the policy for tight insertion, multi-part assembly, and screwing.
Sound on! 🔊
🧵👇
Introducing Play2Perfect: first learn to play with objects, then perfect the policy for tight insertion, multi-part assembly, and screwing.
Sound on! 🔊
🧵👇
10
52
288
68,854
Ritvik Singh retweeted
FPO++ code is out!
We’re releasing the full training and evaluation code for locomotion, motion tracking, and manipulation.
github.com/amazon-far/fpo…
Co-led with @brenthyi
We’re releasing the full training and evaluation code for locomotion, motion tracking, and manipulation.
github.com/amazon-far/fpo…
Co-led with @brenthyi
New project! Flow Policy Gradients for Robot Control
tldr; a simple online RL recipe for training and fine-tuning flow policies for robots
co-led w/ @redstone_hong: hongsukchoi.github.io/fpo-control
0
23
202
17,616
Ritvik Singh retweeted
Important work that goes over the details of training behaviour cloning models with their ABCDataset in collaboration with @xdofai. Lots of interesting details in the appendix on how they went about speeding up the dataloader for reading MP4 files, fast inference to avoid GPU sitting idle and comparing various design choices involved in architectures and evaluation in real and in sim. I enjoyed reading this work.
Introducing ABC: open data, training, and infrastructure for robotics.
We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques.
@arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcall
2
8
44
6,327
Ritvik Singh retweeted
Congratulations on the release!
Proud to have our hardware as part of the foundation this was built on. This is exactly the kind of research that moves the whole field forward — and we're thrilled to play a role in making it possible.
Here's to more breakthroughs ahead. 🤖
@ritvik_singh9 @arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcall
Proud to have our hardware as part of the foundation this was built on. This is exactly the kind of research that moves the whole field forward — and we're thrilled to play a role in making it possible.
Here's to more breakthroughs ahead. 🤖
@ritvik_singh9 @arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcall
Introducing ABC: open data, training, and infrastructure for robotics.
We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques.
@arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcall
0
3
30
4,265
Ritvik Singh retweeted
In computer vision, the field advanced significantly due to openly available datasets and models; hope this will be true in robotics as well.
Introducing ABC: open data, training, and infrastructure for robotics.
We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques.
@arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcall
7
24
238
34,311
Ritvik Singh retweeted
ABC:
- Fully reproducible robotics stack ✅
- Largest open teleoperation dataset ✅
- Top trending dataset on @huggingface ✅
- >400 hrs of sim teleop data, evals dataset with full set of trials, more code coming soon ✅
- Fully reproducible robotics stack ✅
- Largest open teleoperation dataset ✅
- Top trending dataset on @huggingface ✅
- >400 hrs of sim teleop data, evals dataset with full set of trials, more code coming soon ✅
My favorite part: ABC is a reproducible robotics stack. The website demos are not just cool videos: the underlying data, recipes, and infra are open, so people can actually inspect and build on them.
Also excited about the >400h of sim teleop data coming soon. This should help the community iterate faster and study questions like sim-real co-training. One encouraging signal from our experiments: even with only ~2% simulated data mixed into co-training, we observed strong sim-real performance correlations.
Excited to see what the community builds on top!
12
3
45
5,313
Ritvik Singh retweeted
Introducing Do as I Do 👀, a framework to transform everyday human videos into 100s of dexterous robot demos. Co-led with @bhawna_paliwal_ and @HarithejaE, and check out @notmahi's thread!
Here’s a little preview of our dexterous manipulation results. More about how we produce ructions in this mini-thread! 🧵
x.com/notmahi/status…
Here’s a little preview of our dexterous manipulation results. More about how we produce ructions in this mini-thread! 🧵
x.com/notmahi/status…
Robots are the bottleneck in scaling robotics, and learning from human video promises to solve it. But how can chaotic human data ever measure up to sanitized, lab-made teleoperation data?
Introducing Do as I Do: establishing a much needed correspondence between human videos and ome fun insights below: 🧵
8
36
225
28,920
Ritvik Singh retweeted
Excited to share Do as I Do! We turn everyday human videos into physically consistent robot data that can be directly executed in the real world.
This was a fun collaboration with @bhawna_paliwal_ and @willjhliang, with lots of moving parts. More details in Mahi's thread below👇
This was a fun collaboration with @bhawna_paliwal_ and @willjhliang, with lots of moving parts. More details in Mahi's thread below👇
Robots are the bottleneck in scaling robotics, and learning from human video promises to solve it. But how can chaotic human data ever measure up to sanitized, lab-made teleoperation data?
Introducing Do as I Do: establishing a much needed correspondence between human videos and ome fun insights below: 🧵
4
12
32
4,878
Ritvik Singh retweeted
Watching a policy that just 'gets' the task is surreal. All from scaling up the simplest objective there is
Introducing ABC: open data, training, and infrastructure for robotics.
We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques.
@arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcall
3
17
94
14,810
Ritvik Singh retweeted
My favorite part: ABC is a reproducible robotics stack. The website demos are not just cool videos: the underlying data, recipes, and infra are open, so people can actually inspect and build on them.
Also excited about the >400h of sim teleop data coming soon. This should help the community iterate faster and study questions like sim-real co-training. One encouraging signal from our experiments: even with only ~2% simulated data mixed into co-training, we observed strong sim-real performance correlations.
Excited to see what the community builds on top!
Also excited about the >400h of sim teleop data coming soon. This should help the community iterate faster and study questions like sim-real co-training. One encouraging signal from our experiments: even with only ~2% simulated data mixed into co-training, we observed strong sim-real performance correlations.
Excited to see what the community builds on top!
Introducing ABC: open data, training, and infrastructure for robotics.
We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques.
@arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcall
0
9
80
15,808
Ritvik Singh retweeted
Everything we do in robotics and machine learning is built on open source: so cool to see the modern imitation learning stack being released for us to build on :)
Introducing ABC: open data, training, and infrastructure for robotics.
We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques.
@arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcall
0
2
10
1,304
Ritvik Singh retweeted
Very impressive effort, congrats on the team!!
Introducing ABC: open data, training, and infrastructure for robotics.
We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques.
@arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcall
1
2
13
2,105
Ritvik Singh retweeted
also check out our code here github.com/amazon-far/abc
we haven't released everything yet but there's enough there that you can train a policy, eval it in sim, and view the rollouts live in viser!
we haven't released everything yet but there's enough there that you can train a policy, eval it in sim, and view the rollouts live in viser!
2
5
54
15,787
Ritvik Singh retweeted
Amazing effort! Open-sourcing 3,500 hours of data will surely enable a lot of meaningful research in the field. Congrats to the team!
Introducing ABC: open data, training, and infrastructure for robotics.
We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques.
@arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcall
2
5
90
15,251
Ritvik Singh retweeted
So excited for this to finally be out:
A fully open-source data & stack released so we can learn the ABCs of bimanual manipulation together!
This was a huge team effort, and many long days and nights of work. Congrats to the amazing team, I'm very proud of you!
A fully open-source data & stack released so we can learn the ABCs of bimanual manipulation together!
This was a huge team effort, and many long days and nights of work. Congrats to the amazing team, I'm very proud of you!
Introducing ABC: open data, training, and infrastructure for robotics.
We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques.
@arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcall
4
21
182
30,987
Ritvik Singh retweeted
really excited to have got this out!
here's an uncut, real-time video of our policy folding a cardboard box
here's an uncut, real-time video of our policy folding a cardboard box
Introducing ABC: open data, training, and infrastructure for robotics.
We release the largest teleop dataset to date, and extensively investigate design decisions, pretraining, and post-training techniques.
@arthurallshire @Cinnabar233 @adamrasb @redstone_hong @davidrmcall
10
18
239
30,640