SciForma：科学图表的结构忠实生成
SciForma: Structure-Faithful Generation of Scientific Diagrams
浏览论文内容
中文总结 AI 辅助
研究针对科学图表结构保真度问题，提出SciForma框架，分解图表质量为三个结构轴，用M-DPO优化，策划训练和评估数据，实现迭代编辑，使SciForma-9B超越开源基线和GPT-Image-1.5，提升科学图表生成的结构保真度。
中文摘要 AI 辅助
结构保真度对于科学方法图表至关重要。为传达研究逻辑，此类图表必须如实呈现组件、方向关系和文本注释。由于单个错误（如箭头反向或方程式不可读）会使整个图表无效，结构保真度本质上是联合性的。当前开源模型不满足此标准。监督微调（SFT）学习合理布局但无法可靠确保结构正确性，基于标量奖励的训练后处理会掩盖失败的结构维度。为解决此问题，我们引入SciForma框架。它将图表质量分解为组件、箭头和文本三个结构轴，基于此策划了用于结构化训练的SciFormaData-700K和用于逻辑验证评估的SciFormaBench-2K。我们开发了多维联合偏好优化（M-DPO），在训练后跨所有轴强制同时正确，并将梯度自适应路由到最不足的维度。推理时可进行迭代编辑以纠正残留错误。这使得SciForma-9B在SciFormaBench-2K和AIBench上超越所有开源基线和GPT-Image-1.5，使开放科学图表生成接近专有级别的结构保真度。
英文摘要
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.
发表机构
- Wangxuan Institute of Computer Technology, Peking University（北京大学王选计算机技术研究所）
- State Key Lab of CAD & CG, Zhejiang University（浙江大学CAD&CG国家重点实验室）
- Microsoft Research Asia（微软亚洲研究院）
机构由 AI 辅助整理，请以论文原文为准。
