RLWRLD says five of its papers were accepted to CoRL 2026
RLWRLD describes HuRo as a pipeline that turns human videos into robot-aligned observations and actions, scaled to 630,000 episodes for pretraining vision-language-action models.
TLDR
RLWRLD announced five paper acceptances for CoRL 2026. It highlights HuRo, a pipeline it says converts human videos into robot-aligned training data, and Modular Sensory Stream, which it describes as incorporating touch and torque feedback into vision-language-action models. Other papers it describes tackle collision-free path planning and separate robot geometry and action realization from world models so those models can focus on learning how objects respond.
Combined views
3.4K
2 Sources, first seen 19d ago