S1-Omni
科学多模态统一理解与生成 Unified scientific multimodal intelligence
磐石 ScienceOne 团队研发的科学多模态统一理解和生成模型，面向文本、图像、化学材料、蛋白质与谱图等复杂科学对象，建立统一表示、推理与生成能力。
A unified scientific multimodal understanding and generation model from ScienceOne, built for text, images, chemistry, materials, proteins, spectra, and the reasoning paths between them.
S1-Omni 将科学对象转化为可共享的隐空间表示，让理解、预测、生成和编辑形成一个连续工作流。
S1-Omni maps scientific objects into shared hidden representations, connecting understanding, prediction, generation, and editing in one workflow.
面向论文图表、实验图像、复杂示意图与自然语言问题，支持跨模态解析、细粒度问答、过程推理与结构化摘要。
Parse papers, figures, experimental images, diagrams, and natural-language questions with cross-modal reasoning and structured answers.
根据科学语义、实验条件与图文上下文生成图像，也可对局部区域进行编辑、重绘与风格一致的修复。
Generate images from scientific semantics, conditions, and context, then edit, repaint, and restore regions with visual consistency.
统一处理分子式、结构、SMILES 与材料描述，输出面向筛选、优化和实验设计的多维性质预测。
Unify formulas, structures, SMILES, and material descriptions to predict properties for screening, optimization, and experimental design.
从序列、功能描述与多模态上下文中推断结构线索，为蛋白质设计、功能分析与下游模拟提供起点。
Infer structural signals from sequences, functional descriptions, and multimodal context for design, analysis, and downstream simulation.
利用 NMR、MS、IR 等谱图模式进行分子结构推断，并输出可解释的候选结构、关键峰归因与预测值。
Use NMR, MS, IR, and related spectral patterns to infer molecular structures with interpretable candidates, peak attributions, and predicted values.