m2sv: A Scalable Benchmark for Map-to-Street-View Spatial Reasoning

Yosub Shin, Michael Buriek, Igor Molybog · 2026

Read m2sv: A Scalable Benchmark for Map-to-Street-View Spatial Reasoning by Yosub Shin, Michael Buriek, Igor Molybog (2026) 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-2601-19099.txt.

Description

Yosub Shin, Michael Buriek, and Igor Molybog test whether visual models can match a map to a street photo. Vision language models ace many picture tests, yet they stay shaky on spatial tasks that link a top down map to a ground level view. Their bench, m2sv, poses exactly that match as a task.

Each item pairs a north up overhead map with a street view shot taken at the same real place. The model must infer which way the camera faced. Skill at this match matters for robots and aids that must tie what they see to where they are on a map.

Catalog record

CREATORYosub Shin, Michael Buriek, Igor Molybog
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-19099

Browse the full library · About the Tower of Babel