Embodied Spatial Intelligence: From 3D Perception to 3D Reasoning

Jiading Fang · 2025

Read Embodied Spatial Intelligence: From 3D Perception to 3D Reasoning by Jiading Fang (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-00465.txt.

Introduction

Jiading Fang's doctoral thesis sets out embodied spatial intelligence, the aim of giving robots a grasp of space so they can act on spoken instructions in the real world. Large language models handle words well but lack a direct tie to physical places.

Fang works on two fronts. For perception, he builds scene models with implicit neural methods, including self-taught camera calibration, detailed depth fields, and large-scale rebuilding of scenes. For reasoning, he adds a navigation test set, a way to ground language, meaning link words to places, in three dimensions, and a feedback method for long tasks. The thesis lays groundwork for robots that see their surroundings and follow complex commands.

Catalog record

CREATORJiading Fang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE229 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-00465

Browse the full library · About the Tower of Babel