Urban In-Context Learning: Bridging Pretraining and Inference through Masked Diffusion for Urban Profiling

Ruixing Zhang, Bo Wang, Tongyu Zhu, Leilei Sun, Weifeng Lv · 2025

Read Urban In-Context Learning: Bridging Pretraining and Inference through Masked Diffusion for Urban Profiling by Ruixing Zhang, Bo Wang, Tongyu Zhu, Leilei Sun, Weifeng Lv (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-2508-03042.txt.

Description

Ruixing Zhang and co-authors wrote a 2025 paper on urban profiling, predicting how city districts score on social and economic gauges, including places with no data. Older methods learn area features first, then bolt on a simple predictor. Their one stage design borrows the in context trick of language models.

A masked diffusion transformer treats each district's profile as a distribution, not a fixed number. An alignment step steadies training by tying inner features to classic methods' outputs. On three gauges across two cities, the single stage beat the two stage leaders. Parts tests credit the diffusion core most.

Catalog record

CREATORRuixing Zhang, Bo Wang, Tongyu Zhu, Leilei Sun, Weifeng Lv
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE45 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-03042

Browse the full library · About the Tower of Babel