Are we using the best representations for OfflineRL? We found that using latent diffusion models work better at capturing the complex multi-modal distribution of Q-values in Offline RL datasets. Learn about the details from Siddarth Venkatraman tomorrow at ICLR 2026 4:30 in Halle B #157

Daily Training of Robots Driven by RL Segments of daily training for robots driven by reinforcement learning. Multiple tests done in advance for friendly service humans.😊 The training includes some extreme tests, please do not imitate. #AI #Unitree #AGI #EmbodiedIntelligence

This morning, Nature published two papers on bridge editing, the new genome engineering technology from @ArcInstitute: nature.com/articles/s4158…, nature.com/articles/s4158…. I'm quite excited about its potential! Since the whole thing is pretty arcane, I fed the blog post

we've shown MoEs help deep RL agents, but what if we turn up non-stationarity to 11 with multi-task and continual RL? We explore this in our paper, led by Timon Willi & Johan Obando-Ceron 👍🏽 , & w/ Jakob Foerster & Gintare Karolina Dziugaite , accepted RL_Conference ! paper: arxiv.org/abs/2406.18420 1/8

Popular information nobelprize.org/prizes/physics…

nGPT: Normalized Transformer with Representation Learning on the Hypersphere. arxiv.org/abs/2410.01131. Remarkable efficient. (reducing the number of training steps required to achieve the same accuracy by a factor of 4 to 20)

import brain brain.loading("executive_function") # DEBUG: Insufficient sleep detected. Retrying... # DEBUG: Compensatory mechanisms activated (Efficiency -20%) brain.run("today_tasks") ....

🚀 Exciting Opportunity! 🚀 linkedin.com/posts/mariano-…

Please robot… I’m out of toilet paper. No yelling. No awkward moments. Just a smooth, silent rescue. 🧻🤖😂 #robotics #AI #robotsdoingthings #vla #funrobotics

RL is not all you need, nor attention nor Bayesianism nor free energy minimisation, nor an age of first person experience. Such statements are propaganda. You need thousands of people working hard on data pipelines, scaling infrastructure, HPC, apps with feedback to drive

Read more in our blog post, including early notes from large-scale ablations on our pretraining data. Blog: generalistai.com/blog/nov-04-20…

After 18 months in stealth, dozens of prototypes, millions of real-home demonstrations, and one final all-nighter, we’re thrilled for you to say hello to Memo

What if a humanoid robot could choose how to interact with the environment 🤖 — soft when it needs compliance, stiff when it needs precision, and force-aware when it must push/pull? That’s exactly what our Heterogeneous Meta-Control (HMC) framework enables. Our new framework

"We do this not because it is easy, but because we thought it would be easy" Sunday 😂