让科学图示生成精准还原组件、箭头和文字关系，错误零容忍。
SciForma: Structure-Faithful Generation of Scientific Diagrams
- 按组件、箭头、文字三维度拆解图示质量，确保每部分都正确
- 在70万数据上训练，2千个逻辑验证样本评估，性能超越现有开源模型
- 支持生成后迭代纠错，适合科研人员制作高精度论文图表
科学方法图示的结构保真至关重要。为准确传达研究逻辑，图示必须忠实呈现组件、方向关系和文本注释。任何单一错误（如箭头反向或方程不可读）都会使整个图示失效，因此结构保真具有强连贯性：任一维度出错无法被其他维度弥补。当前开源模型无法满足此要求。监督微调（SFT）虽能生成合理布局，但难以保证结构正确；基于标量奖励的后训练则无法定位具体失败维度。为此，我们提出SciForma框架，实现科学方法图示的结构保真生成。具体地，SciForma将图示质量分解为组件、箭头、文本三个结构轴，依托结构清单进行建模。在此基础上，我们构建了包含70万样本的SciFormaData-700K用于结构化训练，以及包含2千个逻辑验证样本的SciFormaBench-2K用于评估。为弥补SFT的不足，我们设计多维连贯偏好优化（M-DPO），在后训练中强制所有维度同时正确，并自适应将梯度导向最薄弱维度。相同的结构清单也支持推理时的迭代编辑以修正残余错误。该组合使SciForma-9B在SciFormaBench-2K与AIBench上均超越所有开源基线及GPT-Image-1.5，使开源科学图示生成接近专有水平的结构保真度。代码与数据将公开于https://github.com/microsoft/SciForma。
原文摘要 · Abstract (English)
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.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。