E²DT hits 86.4% terminal‑bench

- A new Decision Transformer variant called E²DT reached 86.4% on the Terminal‑Bench benchmark, showing improved long‑horizon control in simulated tasks and sample efficiency. - Other robotics papers this cycle include 'Robotic World Model' for policy optimisation, 'Stereo Multistage Spatial Attention' for mobile manipulation, and the Lucid‑XR data engine. - These results provide startups clearer blueprints for robust perception, control, and data infrastructure in physical robots. (x.com 1) (x.com 2)