Rose Hendrix

robotics research at @allen_ai

[Rose Hendrix @rosemhendrix]

@ainaz_eftekhar @ab_deshpande @mayasguru @SnehalJauhri @shahdhruv_ @RanjayKrishna MolmoB0T: https://t.co/3kpPS97dU9

RING: https://t.co/MDolb3V86H

[Rose Hendrix @rosemhendrix]

And that's a wrap on a fantastic ICRA 2026! 🎉 Incredible run for MolmoBot — clean sweep on workshops, winning Best Paper at all three we entered: Synthetic Data for Robot Learning, Beyond Teleoperation, and VLA Pipelines. 🤖

[Rose Hendrix @rosemhendrix]

@ainaz_eftekhar @ab_deshpande @mayasguru @SnehalJauhri @shahdhruv_ @RanjayKrishna Thanks to all the organizers and everyone who stopped by the poster sessions; I had excellent conversations all week. Oh and the gala might be the best I've ever been to. The imperial apartments and Spanish Riding School at the Hofburg?? I am ruined for all future galas 😭✨

[Rose Hendrix @rosemhendrix]

Our paper MolmoB0T won top honor at the SDRL workshop today! Congrats to my co-authors @ab_deshpande, Maya, Snehal, @RanjayKrishna, @shahdhruv_ (plus others, you know how it goes), and we'll see you as well at the VLA Pipelines and Beyond Teleoperation workshops on Friday 🚀

[Abhay Deshpande @ab_deshpande]

I had an "oh shit" moment the first time I ran pi0 - we're now in an era where you can just download a manipulation policy, put it on your robot, and watch things happen. By releasing MolmoBot, now I hope I can help others have their own "oh shit" moments!

[Rose Hendrix @rosemhendrix]

One thing I'm proud of: we release unedited agent-view videos for every rollout in our real-world evaluation set. Not the most riveting viewing! But this kind of transparency helps people understand what policies actually can and can't do. I'd love to see more projects do this.

[Rose Hendrix @rosemhendrix]

Delighted to be releasing our entire system today in support of open robotics research! Now anyone can use MolmoBot, train on our data, or generate their own data to do zero-shot sim-to-real transfer on any robot. Check it out! https://t.co/3kpPS97LJH

[Ai2 @allen_ai]

MolmoBot, our open robotic manipulation suite trained entirely in simulation, now has code, training data, a data generation pipeline, & evals all available.

This puts our robotics models within reach of any research lab—no extensive real-world data collection required. 🧵

[Rose Hendrix @rosemhendrix]

Everything is open: code, data, the whole pipeline. Can't wait to see what this unlocks in robotics research and what people build.

Check it out!
General: https://t.co/JeeSuh4yrD
Technical: https://t.co/v4k1D0qNJj
Code and data coming soon.

[Rose Hendrix @rosemhendrix]

So proud to share MolmoBot! We trained static and mobile manipulation policies entirely in simulation—no real-world teleoperation needed—and they outperform models trained on proprietary data. 🧵

[Ai2 @allen_ai]

Today, a step forward in open robotics - our results show that sim-to-real zero shot transfer for manipulation is possible. MolmoBot is our open model suite for robotics, trained entirely in simulation on MolmoSpaces.🧵

[Rose Hendrix @rosemhendrix]

We use no task-specific fine-tuning or real data at all and were able to evaluate zero-shot in diverse settings, including an external eval from our collaborators.

[Rose Hendrix @rosemhendrix]

So proud to finally share MolmoSpaces with the world! We built this to enable true zero-shot benchmarking for embodied AI—with massive-scale environments to push generalization. Stay tuned for what this unlocks for policy training 🚀

Check it out at https://t.co/gEENwA23xP

[Ai2 @allen_ai]

Introducing MolmoSpaces, a large-scale, fully open platform + benchmark for embodied AI research. 🤖

230k+ indoor scenes, 130k+ object models, & 42M annotated robotic grasps—all in one ecosystem.

[Abhay Deshpande @ab_deshpande]

How should a robot hold a water bottle? 🤔 That depends: is it opening it, or passing it to you?

I’m excited to introduce GraspMolmo, a VLM that predicts semantically appropriate grasps based on your command!

Website: https://t.co/JM83W3nS2g

🧵 Thread ↓

[Rose Hendrix @rosemhendrix]

I'm not making it out to ICRA this year, but come check out Jiaheng's excellent work with Ai2!

[Jiaheng Hu @JiahengHu1]

Excited to be in ATL for #ICRA2025 to present 🔥FLaRe: fine-tuning large transformer policies with #RL, 15:25 Tuesday @ room 410!

I will also be attending the 📷Doctoral Consortium on Monday to talk about my research on self-improving robots. Happy to meet old and new friends!

[Prior @ AI2 @Ai2Prior]

📢Applications are open for summer'25 internships at the PRIOR (computer vision) team
@allen_ai:

Come join us in building large-scale models for:
📸 Open-source Vision-Language Models
💻 Multimodal Web Agents
🤖 Embodied AI + Robotics
🌎 Planet Monitoring

Apply by December 11, 2024!

[Rose Hendrix @rosemhendrix]

Poster ID 61 in the #CoRL2024 4pm poster session, come talk to us about on-policy RL with transformers at scale for sim-to-real visual navigation!

[Kuo-Hao Zeng @KuoHaoZeng]

🚀 Quick Update 🚀

🎉 @ZCCZHANG will present PoliFormer at CoRL Oral Session 5 (🕤 9:30-10:30, Fri, Nov 8, CET)! 🎉

Meet us at Poster Session 4 (🕓 16:00-17:30) to chat with @ZCCZHANG, @rosemhendrix, and Jordi!

💻 Our code & checkpoints are NOW public: https://t.co/Qpkz8sVO3P