SheetDesigner: MLLM-Powered Spreadsheet Layout Generation with Rule-Based and Vision-Based Reflection

Qin Chen, Yuanyi Ren, Xiaojun Ma, Mugeng Liu, Han Shi, Dongmei Zhang · 2025

Read SheetDesigner: MLLM-Powered Spreadsheet Layout Generation with Rule-Based and Vision-Based Reflection by Qin Chen, Yuanyi Ren, Xiaojun Ma, Mugeng Liu, Han Shi, Dongmei Zhang (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-2509-07473.txt.

Description

Qin Chen and 5 co-authors present SheetDesigner in 2025, a system that lays out spreadsheets automatically with a multimodal language model. Spreadsheet layout is skilled manual work: components must sit on a discrete grid, not at free coordinates, and cells carry semantic links such as data dependencies that generic layout tools ignore. The authors first formalize the layout generation task itself.

They supply a seven-criterion evaluation protocol and a dataset of 3,326 spreadsheets. SheetDesigner then works zero-shot, with no training, combining rule-based reflection and vision-based reflection to place components and fill content. It beats five baselines by at least 22.6 percent. The study also finds that models handle overlap and balance well through vision but struggle with alignment, which is why both reflection types are needed. Code and data are public.

Catalog record

CREATORQin Chen, Yuanyi Ren, Xiaojun Ma, Mugeng Liu, Han Shi, Dongmei Zhang
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE63 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2509-07473

Browse the full library · About the Tower of Babel