XDOF (@xdofai) | Vanlett

XDOF reposted

@uynitsuj

Jun 30

Introducing WARP-RM: A Warp-Augmented Relative Progress Reward Model for Data Curation 🧵 We gave our t-shirt folding robot more demonstrations and it got worse. Every extra demo ended in a successfully folded shirt. The data wasn't bad. It was noisy. The policy couldn't tell productive motion from dead time, and it imitated both equally. So which moments of a demo are actually worth copying? 🌐 Project Website: uynitsuj.github.io/warp-rm 📄 Paper: arxiv.org/abs/2606.28320 💻 Code: github.com/uynitsuj/WARP-RM 📨 XDOF blog post: xdof.ai/blog/warp-rm

0:50

6

22

141

26,794

XDOF

@xdofai

Jun 30

Read more about WARP-RM on the XDOF blog: xdof.ai/blog/warp-rm Check out the WARP-RM project page: uynitsuj.github.io/warp-rm/

Lowering the Noise Floor on Robot Data

We gave a folding robot more demonstrations and it got worse. Chasing down why led to a way of teaching it to tell productive moments from dead time, with no labels.

xdof.ai

4

300

XDOF

@xdofai

Jun 30

Why did giving our robot more demonstrations make it worse? Introducing WARP-RM: a self-supervised reward model that learns which moments of a demo are actually worth imitating. Result: 1.8x the throughput compared to vanilla BC on t-shirt folding! Congrats to lead authors: @uynitsuj (XDOF intern) Andrew Goldberg @kavish_kondap Karim El-Refai (XDOF) Ethan Ransing @QianzhongChen (XDOF intern)

2

5

43

3,299

XDOF

@xdofai

Jun 23

Read more about SARM2 on the XDOF blog: xdof.ai/blog/sarm2 Check out the SARM2 project page: qianzhong-chen.github.io/sar…

The Robot That Learned to Grade Itself

Reward models are reshaping how machines learn to manipulate the world. SARM2 can grade dozens of tasks and turn raw demonstrations into policies that keep improving on their own.

xdof.ai

3

8

716

XDOF

@xdofai

Jun 23

SARM2 is a multi-task, stage-aware reward model: dense, accurate AND general. Feeding SARM2 rewards into a self-improvement framework SPIRAL boosts task success rates from 50-60% to 90-100% 🔥 Congrats to @QianzhongChen (XDOF intern) and co-authors: - @BrianZheng103 - @uynitsuj (XDOF intern) - @suning_huang - @JiankaiSun

4

8

58

7,843

XDOF

@xdofai

Jun 18

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 ✅

Guanya Shi

@GuanyaShi

Jun 18

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

4,125

XDOF reposted

Ritvik Singh

@ritvik_singh9

Jun 17

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

2:52

29

106

676

300,230

XDOF

@xdofai

Jun 17

Read more and access the data here: xdof.ai/blog/abc-130k

ABC-130K: The largest open source teleoperation dataset

130K episodes, 200 complex manipulation tasks. All on a low-cost bimanual rig. Fully open source under Apache 2.0.

xdof.ai

2

3

8

2,094

XDOF

@xdofai

Jun 17

We are proud to release ABC-130K, the largest open bimanual manipulation dataset. ABC-130K includes 130K trajectories covering 200 complex manipulation tasks. All are performed on a low-cost bimanual rig (YAM by I2RT). The data is available under the highly permissive Apache 2.0 license.

11

22

102

18,729

XDOF

@xdofai

May 21

luma.com/qxgpdp6y

Sunset cruise on the Danube with Bessemer & XDOF · Luma

Join Bessemer Venture Partners and XDOF for a sunset cruise down the Danube during ICRA week. We're bringing together a small group of robotics founders,…

luma.com

1

4

1,225

XDOF

@xdofai

May 21

Headed to ICRA? We're hanging out with our friends at @BessemerVP on a boat, and you should join us! Come enjoy golden hour on the Danube with a small group of robotics researchers, founders and operators on Tuesday, June 2nd from 7pm-9pm. Link below 🔽

2

4

19

117,434