SO-Bench: A Structural Output Evaluation of Multimodal LLMs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Feng, Di, Ma, Kaixin, Nan, Feng, Chen, Haofeng, Zhai, Bohan, Griffiths, David, Gao, Mingfei, Gan, Zhe, Verma, Eshan, Yang, Yinfei, Chen, Zhifeng, Dehghan, Afshin
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915870467424256
author Feng, Di
Ma, Kaixin
Nan, Feng
Chen, Haofeng
Zhai, Bohan
Griffiths, David
Gao, Mingfei
Gan, Zhe
Verma, Eshan
Yang, Yinfei
Chen, Zhifeng
Dehghan, Afshin
author_facet Feng, Di
Ma, Kaixin
Nan, Feng
Chen, Haofeng
Zhai, Bohan
Griffiths, David
Gao, Mingfei
Gan, Zhe
Verma, Eshan
Yang, Yinfei
Chen, Zhifeng
Dehghan, Afshin
contents Multimodal large language models (MLLMs) are increasingly deployed in real-world, agentic settings where outputs must not only be correct, but also conform to predefined data schemas. Despite recent progress in structured generation in textual domain, there is still no benchmark that systematically evaluates schema-grounded information extraction and reasoning over visual inputs. In this work, we conduct a comprehensive study of visual structural output capabilities for MLLMs with our carefully designed SO-Bench benchmark. Covering four visual domains, including UI screens, natural images, documents, and charts, SO-Bench is built from over 6.5K diverse JSON schemas and 1.8K curated image-schema pairs with human-verified quality. Benchmarking experiments on open-sourced and frontier proprietary models reveal persistent gaps in predicting accurate, schema compliant outputs, highlighting the need for better multimodal structured reasoning. Beyond benchmarking, we further conduct training experiments to largely improve the model's structured output capability. We make the benchmark and evaluation publicly available at https://github.com/apple/ml-sobench
format Preprint
id arxiv_https___arxiv_org_abs_2511_21750
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SO-Bench: A Structural Output Evaluation of Multimodal LLMs
Feng, Di
Ma, Kaixin
Nan, Feng
Chen, Haofeng
Zhai, Bohan
Griffiths, David
Gao, Mingfei
Gan, Zhe
Verma, Eshan
Yang, Yinfei
Chen, Zhifeng
Dehghan, Afshin
Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
Robotics
Multimodal large language models (MLLMs) are increasingly deployed in real-world, agentic settings where outputs must not only be correct, but also conform to predefined data schemas. Despite recent progress in structured generation in textual domain, there is still no benchmark that systematically evaluates schema-grounded information extraction and reasoning over visual inputs. In this work, we conduct a comprehensive study of visual structural output capabilities for MLLMs with our carefully designed SO-Bench benchmark. Covering four visual domains, including UI screens, natural images, documents, and charts, SO-Bench is built from over 6.5K diverse JSON schemas and 1.8K curated image-schema pairs with human-verified quality. Benchmarking experiments on open-sourced and frontier proprietary models reveal persistent gaps in predicting accurate, schema compliant outputs, highlighting the need for better multimodal structured reasoning. Beyond benchmarking, we further conduct training experiments to largely improve the model's structured output capability. We make the benchmark and evaluation publicly available at https://github.com/apple/ml-sobench
title SO-Bench: A Structural Output Evaluation of Multimodal LLMs
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
Robotics
url https://arxiv.org/abs/2511.21750