TopoAlign : Topology-Aware Visual Representation Alignment

\authororcid Xinyuan Yan0000-0003-3396-1310, \authororcid Rita Sevastjanova0000-0002-2629-9579, \authororcid Mennatallah El-Assady0000-0001-8526-2613, and \authororcid Bei Wang0000-0002-9240-0700 · 2026

Read TopoAlign : Topology-Aware Visual Representation Alignment by \authororcid Xinyuan Yan0000-0003-3396-1310, \authororcid Rita Sevastjanova0000-0002-2629-9579, \authororcid Mennatallah El-Assady0000-0001-8526-2613, and \authororcid Bei Wang0000-0002-9240-0700 (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-2605-25541.txt.

Description

Xinyuan Yan and colleagues compare neural representations by shape rather than by local geometry alone. Existing alignment scores lean on neighborhood and cluster overlap, which reveal little about global layout. TopoAlign builds mapper graphs from shared inputs, aligns them with joint force-directed layouts, spots matching regions with Bubble Sets, and probes fine patterns with motif queries. Case studies on language and multimodal models, plus expert feedback, show the tool surfaces structural insight that geometry-only scores miss.

Catalog record

CREATOR\authororcid Xinyuan Yan0000-0003-3396-1310, \authororcid Rita Sevastjanova0000-0002-2629-9579, \authororcid Mennatallah El-Assady0000-0001-8526-2613, and \authororcid Bei Wang0000-0002-9240-0700
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE73001
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2605.25541

Browse the full library · About the Tower of Babel