PanoWorld: Towards Spatial Supersensing in 360$^\circ$ Panorama World

2026-05-13

AI Summary

Existing multimodal large language models struggle to effectively comprehend the spatial structure of 360-degree panoramic environments and lack holistic spherical scene understanding. This work proposes a native panoramic understanding framework, formally defining its core capabilities for the first time and introducing instruction-tuning data and diagnostic benchmarks that incorporate geometric awareness, language alignment, and depth perception. By explicitly injecting spherical geometric priors, we design a spherical spatial cross-attention mechanism that enables observer-centric continuous spatial reasoning under equirectangular projection. Experiments demonstrate that our approach significantly outperforms both open-source and closed-source models on multiple benchmarks, including PanoSpace-Bench, H* Bench, and R2R-CE Val-Unseen, thereby validating the efficacy of native panoramic modeling.

Abstract

Multimodal large laboratory models (MLLMs) still struggle with spatial understanding under the dominant perspective-image paradigm, which inherits the narrow field of view of human-like perception. For navigation, robotic search, and 3D scene understanding, 360-degree panoramic sensing offers a form of supersensing by capturing the entire surrounding environment at once. However, existing MLLM pipelines typically decompose panoramas into multiple perspective views, leaving the spherical structure of equirectangular projection (ERP) largely implicit. In this paper, we study pano-native understanding, which requires an MLLM to reason over an ERP panorama as a continuous, observer-centered space. To this end, we first define the key abilities for pano-native understanding, including semantic anchoring, spherical localization, reference-frame transformation, and depth-aware 3D spatial reasoning. We then build a large-scale metadata construction pipeline that converts mixed-source ERP panoramas into geometry-aware, language-grounded, and depth-aware supervision, and instantiate these signals as capability-aligned instruction tuning data. On the model side, we introduce PanoWorld with Spherical Spatial Cross-Attention, which injects spherical geometry into the visual stream. We further construct PanoSpace-Bench, a diagnostic benchmark for evaluating ERP-native spatial reasoning. Experiments show that PanoWorld substantially outperforms both proprietary and open-source baselines on PanoSpace-Bench, H* Bench, and R2R-CE Val-Unseen benchmarks. These results demonstrate that robust panoramic reasoning requires dedicated pano-native supervision and geometry-aware model adaptation. All source code and proposed data will be publicly released.

Problem

• panoramic understanding
• spatial reasoning
• equirectangular projection
• spherical geometry
• 360-degree vision

Innovation

• pano-native understanding
• spherical spatial cross-attention
• equirectangular projection
• 360-degree panorama
• geometry-aware MLLM