Room-to-Room (R-2R)

The Room-to-Room (R-2R) dataset/benchmark contains multimodal environments used to evaluate the performance of embodied agents in navigating and understanding complex scenes through perception and reasoning.

Papers using Room-to-Room (R-2R) (2)

PRISM: Perception Reasoning Interleaved for Sequential Decision Making2026Following Route Instructions using Large Vision-Language Models: A Comparison between Low-level and Panoramic Action Spaces2025