Major highlight was meeting some of the researchers that inspired me: Jim Fan with the hot takes and giving me encouragement to push on my ideas Jeannette Bohg was so kind to take time out and chat with me Lerrel Pinto being the coolest guy at the conference (LOVE DynaMem btw)

Thanks a lot Jim Fan Jeannette Bohg and Lerrel Pinto hanging out with students (in this case from my group). upcoming folks meeting visible and senior researchers is often surprisingly influential in the long run!

Just out of curiosity. Who are some of you guys’ favorite researchers, and why?

This was really well done! Super awesome work Mandi and team!

Robots being able to find objects is super important. This work has an interesting approach on how to get over that hump. Awesome work Tenny and team!

Second most common regret of the dying : I wish I hadn’t I worked so hard. bronnieware.com/blog/regrets-o…

Impressive bimanual manipulation demos! Im starting to see more of an uptick of people working on the speed of IL policies and I’m all for it! This robot may already be able to put items in a toolbox and put together a box faster than me 😅. Awesome work Pete Florence Andy Zeng

Sadly I believe this to be mostly true nowadays. This is a huge issue. It puts an immense amount of pressure on undergrads to publish multiple first author papers to even have a chance at a top uni. Also, where does this leave folks who are coming from different backgrounds?

Robotics data is expensive and slow to collect. Robotics labs and companies spend months just to collect around 10k hours of demonstration data, all while that much video is uploaded to YouTube every 20 minutes. However, none of this video data contains action labels. How can we

Tired of collecting robot data? Wish you could use a YouTube video of a person doing the task you want your robot to do as data? AMPLIFY offers insights and an interesting solution to this problem! Make sure to go check it out!

Tired of slow-moving robots? Want to know how learning-driven robots can move closer to industrial speeds in the real world? Introducing SAIL - a system for speeding up the execution of imitation learning policies up to 3.2x on real robots. A short thread: 1/

Most assistive robots live in labs. We want to change that. FEAST enables care recipients to personalize mealtime assistance in-the-wild, with minimal researcher intervention across diverse in-home scenarios. 🏆 Outstanding Paper & Systems Paper Finalist Robotics: Science and Systems 🧵1/8

Imagine a future where robots are part of our daily lives — How can end users teach robots new tasks by directly showing them, just like teaching another person? 🧵👇

Umm…this is a strange teleoperating system. Also, this looks quite expensive 😅

🥋 We're excited to share judo: a hackable toolbox for sampling-based MPC (SMPC), data collection, and more, designed to make it easier to experiment with high-performance control. Try it: pip install judo-rai

We ran more experiments to better understand “why” diffusion models do better in data-constrained settings than autoregressive. Our findings support the hypothesis that diffusion models benefit from learning over multiple token orderings, which contributes to their robustness and

Self-Questioning Language Models: LLMs that learn to generate their own questions and answers via asymmetric self-play RL. There is no external training data – the only input is a single prompt specifying the topic.

Is it just me, or does GPT-5 come off a bit brash?