Happy to share that ReLo has been accepted to #NeurIPS2023 🎉 Much thanks to the reviewers and my amazing collaborators Somjit Nath Pedro Magalhães Samira E. Kahou. Will be my first, first author paper I'm presenting, so looking forward to it!👀 See you at New Orleans!

Super stoked to join Araya Global to work on language models for robotic manipulation in the real world!

Our paper on "PcLast: Discovering Plannable Continuous Latent States" was accepted to ICML Conference 2024!🎉 Was amazing working with Anurag Koul (अनुराग कौल), Alex Lamb, John Langford and the rest of the team last summer. Check out the thread below to learn more!

Stop by the VLMNM Workshop at ICRA on Friday May 17th to chat about how fine-tuning and prompting affect VLM planners!

Fun evening discussing offline RL at @tokyoaijp. W/ Johannes Ackermann✈️COLM and Takuma Seno(Sony AI) Any excuse to advertise my thesis 😁

Wish I had seen this tweet before I spent a few days rerunning experiments before realising the data was off 🙃You live and you learn

It's funny how often the unreasonable effectiveness of action chunking (h/t Kai Arulkumaran @ALIFE2025) comes up in BC for robot learning.

Every bug fix is a personal gradient update

10/10 would recommend

You don't have to worry about me breaking Franka no mo Kai Arulkumaran @CoRL2025 🫡

Tried zero shot generalisation for robotic manipulation with o3. Gave it demos of pick/place in 3 positions and asked it to place at a new location. o3 complained first and then I said "Think Harder!" And Voila! Joint positions directly from o3 #lerobotworldwidehackathon

Fun weekend at the LeRobot Hackathon. What if you could have a bot to play checkers with? o3 provides the 🧠 🤖 executes the move. Or if you are like me and bad at the game, just have o3 play against itself 😅 #lerobotworldwidehackathon #lerobot #huggingface

Soham seeing all the Cluely funding news last week

"A multi-user multi-robot multi-goal multi-device human-robot interaction manipulation benchmark" is now out in Frontiers! 🧑‍🦰🤖🏁🎮 The Araya Global RL team has been working on HRI software that helps us better study multi-agent (multi-user + multi-robot) collaboration