A benchmark consisting of language-conditioned long-horizon (room-to-room and floor-to-floor navigation) MoMa tasks that are paired with human-collected expert demonstrations from both simulated and real-world environments. - View it on GitHub