SciIR：一個用於科學圖像推理生成的大規模訓練數據集與基準

摘要

雖然文字生成圖像（Text-to-Image, T2I）模型在生成照片級真實視覺內容方面已展現顯著成功，但它們在處理科學圖像所需的嚴謹語義對齊與邏輯推理時仍面臨挑戰。受皮爾士符號學三元組啟發，我們提出了科學圖像推理（Scientific Image Reasoning, SciIR），這是一個用於訓練與評估科學圖像生成的綜合性資源。我們將科學推理形式化為三個核心維度：實體結構（像似符）、科學過程（指示符）與科學定律（規約符）。具體而言，為克服科學圖像生成中訓練數據稀缺的問題，我們精心建構了SciIR-82k，這是一個大型數據集，包含來自前沿出版物的超過八萬張高品質科學圖像-文字對。該數據集根據符號學維度進行層級化組織，並納入了科學推理思維鏈（Scientific Reasoning Chain-of-Thought, Sci-RCoT），以明確建模潛在的視覺邏輯。在評估方面，我們提出了SciIR-Bench，該基準與上述三個符號學層級對齊，並採用原子清單（Atomic Checklist）將結果導向的科學準確性轉化為過程導向、可驗證且細粒度的問題。我們的大量實驗揭示了當前模型在科學推理能力上的顯著缺陷。此外，透過在SciIR-82k數據集上進行微調，我們開發了Qwen-Image-SciIR模型，該模型在SciIR-Bench上取得了顯著提升，最終分數從35%提高至43%，為未來科學圖像生成的發展奠定了堅實基礎。

English

While Text-to-Image (T2I) models have shown remarkable success in generating photorealistic visual content, they still struggle with the rigorous semantic alignment and logical reasoning required for scientific imagery. Inspired by Peirce's Semiotic Triad, we introduce Scientific Image Reasoning (SciIR), a comprehensive resource for training and evaluation of scientific image generation. We formalize scientific reasoning into three core dimensions: Entity Structure (Icon), Scientific Process (Index), and Scientific Law (Symbol). Specifically, to overcome the scarcity of training data in scientific image generation, we elaborately create SciIR-82k, a large-scale dataset containing over 80,000 high-quality scientific image-text pairs from cutting-edge publications. The dataset is hierarchically organized according to the semiotic dimensions and incorporates a Scientific Reasoning Chain-of-Thought (Sci-RCoT) to explicitly model underlying visual logic. For evaluation, we propose SciIR-Bench, which aligns with these three semiotic levels and employs an Atomic Checklist to convert the outcome-oriented scientific accuracy into process-oriented, verifiable, fine-grained questions. Our extensive experiments reveal significant deficiencies in current models' scientific reasoning capabilities. Furthermore, by fine-tuning on the SciIR-82k dataset, we developed the Qwen-Image-SciIR model, which achieves a substantial improvement on the SciIR-Bench, increasing the final score from 35\% to 43\%, laying a solid foundation for future advances in scientific image generation.