Dhruv Shah
971
posts
@shahdhruv_
San Francisco, CA
Joined April 2012
Demos have been my favorite part of
@corl_confover the years and
@GuanyaShiis making sure 2026 will have the most robots by far! Call for Demos is now live, submit proposals to bring your robots to Austin. We also have a new fast-track for accepted papers.
One of my favorite parts of this release is our new on-device policy (GROD2), which incorporates cross-embodiment pre-training to seamlessly learn a generalist policy. This enables specializing to new embodiments in <200 demos without losing the pre-training goodness of Gemini.
00:00
01:32
One brain. For any robot. 🤖 We’re launching Gemini Robotics 2: our next-generation physical AI bringing full body intelligence to humanoids, advanced dexterity, multi-robot teamwork and more.
Your generalist policies might be smarter than you think! When orchestrated with a frontier reasoning model, we extract substantial performance improvements, vastly outperforming the best e2e policies or CaP. lianegalanti.github.io/Pigey/ 🧵:
00:00
00:44
Robot policies can move but can't think. LLMs can think but can't move. So we connected them. Real robot: 16.7% → 97.3% Sim (LIBERO-PRO): 12.8% → 53.3%
Excited to host the next edition of Northeast Robotics Colloquium at
@Princeton@PrincetonAInews! Fall is a great time to be on campus and hang out with all the fun roboticists in the area. Call for papers and Rising Stars is live: nerc2026.github.io
NERC is coming to beautiful
@EPrincetonthis Fall! Please see the CfP for papers and demos, as well as our call for Rising Stars in Robotics. Submissions due August 20 AoE. nerc2026.github.io
What makes a VLM good at real robotic tasks? We present a new Robot Question Answering (RQA) benchmark across domains relevant for robotics and study their failure modes. TLDR: Start with Gemini and Qwen (open), and benchmark on RQA! berkeleyautomation.github.io/robovista/ 🧵:
00:39
🤖How well do today's VLMs actually understand real-world robotics? 👀 Excited to share RoboVista at #RSS2026 — a systematic evaluation and benchmark for VLMs across diverse, real-world robot applications. Website, dataset and paper: berkeleyautomation.github.io/robovista/ Developed by
