Benchmarking Layout-Guided Diffusion Models through Unified Semantic-Spatial Evaluation in Closed and Open Settings

Luca Parolari, Nicla Faccioli, Lamberto Ballan · 2026

Read Benchmarking Layout-Guided Diffusion Models through Unified Semantic-Spatial Evaluation in Closed and Open Settings by Luca Parolari, Nicla Faccioli, Lamberto Ballan (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-2604-25358.txt.

Description

Parolari, Faccioli and Ballan put forth a closed-set test, C-Bench, that splits key skills with varied prompt and layout twists, and an open-set test, O-Bench, for real-world prompts and layouts. They also build one shared scoring rule that joins meaning and place scores into a single mark. With these tests, they rate six layout-led diffusion models on 319,086 images, setting a rank with fine splits for text and layout fit. Fine checks across settings and prompt twists show strengths and limits.

Catalog record

CREATORLuca Parolari, Nicla Faccioli, Lamberto Ballan
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE35072
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.25358

Browse the full library · About the Tower of Babel