I've been compiling papers on Physical AI — the intersection of foundation models and robotics. This covers Vision-Language-Action (VLA) models like RT-2 and π₀, world models (DreamerV3, Genie 2), diffusion policies, deployment and latency researches, cross-embodiment transfer, and safety/alignment for robots.

The field has exploded in the past 18 months. We went from "How about we apply llms to robotics" to having so many different research paths and dimensions to optimize.

Contributions welcome — especially corrections and missing papers.