Thinking in 360°: Humanoid Visual Search in the Wild
Published in CVPR 2026, 2026
This paper proposes humanoid visual search, in which a humanoid agent actively rotates its head to search for objects or paths in an immersive world represented by a 360° panoramic image. We introduce H* Bench, the first benchmark for embodied panoramic visual search that moves beyond household scenes to challenging in-the-wild environments such as transportation hubs, large-scale retail spaces, urban streets, and public institutions, and demonstrate that large multimodal models can be trained to perform human-like active head-rotation search in real-world environments.
Recommended citation: Heyang Yu*, Yinan Han*, Xiangyu Zhang, Baiqiao Yin, Bowen Chang, Xiangyu Han, Xinhao Liu, Jing Zhang, Marco Pavone, Chen Feng, Saining Xie, Yiming Li, "Thinking in 360°: Humanoid Visual Search in the Wild", CVPR 2026
Download Paper