SciFlow-Bench: Evaluating Structure-Aware Scientific Diagram Generation via Inverse Parsing

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: Zhang, Tong, Lin, Honglin, Liu, Zhou, Chen, Chong, Zhang, Wentao
Format: Preprint
Publié: 2026
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866914319276441600
author Zhang, Tong
Lin, Honglin
Liu, Zhou
Chen, Chong
Zhang, Wentao
author_facet Zhang, Tong
Lin, Honglin
Liu, Zhou
Chen, Chong
Zhang, Wentao
contents Scientific diagrams convey explicit structural information, yet modern text-to-image models often produce visually plausible but structurally incorrect results. Existing benchmarks either rely on image-centric or subjective metrics insensitive to structure, or evaluate intermediate symbolic representations rather than final rendered images, leaving pixel-based diagram generation underexplored. We introduce SciFlow-Bench, a structure-first benchmark for evaluating scientific diagram generation directly from pixel-level outputs. Built from real scientific PDFs, SciFlow-Bench pairs each source framework figure with a canonical ground-truth graph and evaluates models as black-box image generators under a closed-loop, round-trip protocol that inverse-parses generated diagram images back into structured graphs for comparison. This design enforces evaluation by structural recoverability rather than visual similarity alone, and is enabled by a hierarchical multi-agent system that coordinates planning, perception, and structural reasoning. Experiments show that preserving structural correctness remains a fundamental challenge, particularly for diagrams with complex topology, underscoring the need for structure-aware evaluation.
format Preprint
id arxiv_https___arxiv_org_abs_2602_09809
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle SciFlow-Bench: Evaluating Structure-Aware Scientific Diagram Generation via Inverse Parsing
Zhang, Tong
Lin, Honglin
Liu, Zhou
Chen, Chong
Zhang, Wentao
Computer Vision and Pattern Recognition
Scientific diagrams convey explicit structural information, yet modern text-to-image models often produce visually plausible but structurally incorrect results. Existing benchmarks either rely on image-centric or subjective metrics insensitive to structure, or evaluate intermediate symbolic representations rather than final rendered images, leaving pixel-based diagram generation underexplored. We introduce SciFlow-Bench, a structure-first benchmark for evaluating scientific diagram generation directly from pixel-level outputs. Built from real scientific PDFs, SciFlow-Bench pairs each source framework figure with a canonical ground-truth graph and evaluates models as black-box image generators under a closed-loop, round-trip protocol that inverse-parses generated diagram images back into structured graphs for comparison. This design enforces evaluation by structural recoverability rather than visual similarity alone, and is enabled by a hierarchical multi-agent system that coordinates planning, perception, and structural reasoning. Experiments show that preserving structural correctness remains a fundamental challenge, particularly for diagrams with complex topology, underscoring the need for structure-aware evaluation.
title SciFlow-Bench: Evaluating Structure-Aware Scientific Diagram Generation via Inverse Parsing
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2602.09809