SciForma: Structure-Faithful Generation of Scientific Diagrams

2026-07-20

AI Summary

Existing open-source models exhibit insufficient structural fidelity when generating scientific methodology diagrams, where minor errors can render entire figures invalid. To address this, this work proposes SciForma, a framework that decomposes diagram quality into three structural dimensions—components, arrows, and text—and employs a structured checklist to guide both training and evaluation. SciForma introduces a novel Multidimensional Direct Preference Optimization (M-DPO) mechanism to jointly optimize all dimensions, complemented by iterative in-context editing during inference to ensure global structural correctness. The study also constructs SciFormaBench-2K, the first logic-validated evaluation benchmark, and SciFormaData-700K, a large-scale dataset. Experiments demonstrate that SciForma-9B significantly outperforms all open-source baselines and GPT-Image-1.5 on both SciFormaBench-2K and AIBench, approaching the structural fidelity of closed-source models.

Abstract

Structural fidelity is essential to scientific methodology diagrams. To communicate research logic, these diagrams must faithfully render components, directional relations, and textual annotations. Since a single error, such as a reversed arrow or an unreadable equation, can invalidate the entire figure, structural fidelity is inherently conjunctive: correctness on one axis cannot compensate for failure on another. Current open-source models fail to satisfy this criterion. Supervised fine-tuning (SFT) learns plausible layouts but cannot reliably ensure structural correctness, while scalar reward-based post-training obscures which structural dimension has failed. To address this, we introduce SciForma, a framework for the structure faithful generation of scientific methodology diagrams. Specifically, SciForma decomposes diagram quality into three structural axes: Component, Arrow, and Text, guided by a structural inventory. Built on this foundation, we curate SciFormaData-700K for structured training and SciFormaBench-2K for logic-verified evaluation. To close the gap left by SFT, we develop Multi-Dimensional Conjunctive Preference Optimization (M-DPO), which enforces simultaneous correctness across all axes and adaptively routes gradients to the most deficient dimension in post-training. The same structural inventory also enables iterative editing at inference time to correct residual errors. This combination allows SciForma-9B to exceed all open-source baselines and GPT-Image-1.5 on both SciFormaBench-2K and AIBench, bringing open scientific diagram generation close to proprietary-level structural fidelity. Our code and data will be available at: https://github.com/microsoft/SciForma.

Problem

• structural fidelity
• scientific diagrams
• component correctness
• directional relations
• textual annotations

Innovation

• Structural Fidelity
• Scientific Diagram Generation
• Multi-Dimensional Conjunctive Preference Optimization
• M-DPO
• Structure-Aware Editing