🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

I'm too excited to wait for Ishika Singh to tweet about our new preprint 😊 LLMs can plan, sort of, using commonsense, but they get steps wrong. Formal planning gets things right but multiagent search spaces are exponential. TwoStep is the best of both! arxiv.org/abs/2403.17246

LLMs can act as planners in simple settings, but struggle in complex ones, e.g. multi-agent, where planner has to track agent-specific states. TwoStep leverages strengths of LLMs & classical planners to produce faster, shorter, and guaranteed plans! glamor-usc.github.io/twostep 🧵👇

I'll be presenting SPRINT at ICRA 2024 next week, come by to learn how to scalably pretrain lang-cond. offline RL! Talk: 1:30-3PM in CC-416 Poster: 4:30-6PM at board #07.03 Also, feel free to message me to chat

We are happy to announce the blog of the Robotics and Autonomous Systems Center at USC! We have wanted to show USC’s robotics research to a broader audience, and we want this blog to bring us closer to that! rasc.usc.edu/news/blog-laun…

How can #robots remember? 🤖 💭 For robots to understand and respond to questions that require complex multi-step reasoning in scenarios over long periods of time, we built ReMEmbR, a retrieval-augmented memory for embodied robots. 👀 Technical deep dive from #NVIDIAResearch

How can robots efficiently learn **new tasks/in new settings**? Introducing EXTRACT: a reinforcement learning (RL) framework that extracts a discrete + continuously parameterized skill library from offline data for efficient RL on new tasks! Accepted to CoRL 2024: 🧵👇

Check out my new USC Robotics blog on how we can leverage human saliency to improve image representations for RL! 👀 I am excited to present this work at IROS in the coming week. Read the full blog post here: rasc.usc.edu/blog/visual-re… Huge thanks to Abrar Anwar and Erdem Bıyık

I gave a talk at University of Washington where I went over our group's recent efforts on robot learning from gaze, comparative language, instructions, and demonstrations (at most two!). It was recorded and uploaded on YouTube so I will share it here :) youtu.be/2qQlOwp2v2A

‼️ Come visit our poster on learning adaptive policies under changing latent dynamics at #NeurIPS2024 🧨! Happy to connect and chat! ⏰Thurs 12/12, 11am-2pm 📍W Ballroom A-D 📰Paper: arxiv.org/abs/2402.15957 Done w/ my advisor Erdem Bıyık and my awesome collaborators at Google!

Check out this poster by Robby Costales at #NeurIPS2024, which shows how quality-diversity can facilitate agent training in simulators with unhelpful representations!

📢Exciting news! Our workshop Human-in-the-Loop Robot Learning: Teaching, Correcting, and Adapting has been accepted to RSS 2025!🤖🎉Join us as we explore how robots can learn from and adapt to human interactions and feedback. 🔗Workshop website: hitl-robot-learning.github.io 🧵👇

We see increasingly capable robot policies everyday. Yet during execution, they often act reasonably but fail to complete tasks, e.g. due to novel scenes or objects. Wouldn't it be nice if we provide a handful of interventions to the robot policies and they could learn from them?

Reward models that help real robots learn new tasks—no new demos needed! ReWiND uses language-guided rewards to train bimanual arms on OOD tasks in 1 hour! Offline-to-online, lang-conditioned, visual RL on action-chunked transformers. 🧵

How can non-experts quickly teach robots a variety of tasks? Introducing HAND ✋, a simple, time-efficient method of training robots! Using just a **single hand demo**, HAND learns manipulation tasks in under **4 minutes**! 🧵

Super interesting to see a single demo from a *HUMAN* can provide sufficient supervision to learn robot behaviors! Congrats to the team!

Some play data + 1 human demonstration (without the robot!!) achieves the task. Thread and paper below 👇🏻

Check out our latest work on robot learning from human hand demonstrations! HAND ✋ can learn new manipulation policies in under 4 minutes from a single hand demo! Huge shoutout to my collaborators and Matthew Hong who will be applying to PhD programs this year!

[Blog Post Announcement] The internet is full of “interesting” data: cat videos, think pieces, and highlight reels—but robots often need to learn from mundane data to help us with everyday unexciting tasks. While people aren’t incentivized to share this boring data, we constantly

Just a small reminder that our workshop is happening tomorrow, and we have an amazing line of speakers! Make sure to check out the workshop website for the schedule. 🤖