HoneyBee：面向视觉语言推理器的数据配方
HoneyBee: Data Recipes for Vision-Language Reasoners
论文系统研究了视觉语言推理数据集的构建策略，包括上下文来源、数据干预和维度扩展，并提出了测试时扩展方法，可直接用于构建高质量多模态推理训练数据。
原文依据 we introduce several data curation approaches and study their impacts on VL reasoning capabilities
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论近期视觉语言模型（VLM）的进展使其在推理任务上表现出色。然而，构建高性能视觉语言推理训练数据集的基本原则仍未被充分理解。在本工作中，我们引入了若干数据管理方法，并通过精心控制训练和评估设置来研究它们对视觉语言推理能力的影响。我们分析了上下文（图像和问题对）来源的影响，实施了针对性的数据干预，并探索了图像、问题和思维链（CoT）解决方案的扩展。我们的发现表明：（a）上下文来源策略显著影响VLM性能；（b）如图像描述辅助信号和引入纯文本推理等干预措施带来了显著增益；（c）扩展所有数据维度（例如每张图像的独特问题数和每个图像-问题对的独特CoT数）能持续提升推理能力。基于这些洞见，我们推出了HoneyBee，一个大规模、高质量的CoT推理数据集，包含250万个示例，由35万个图像-问题对组成。使用HoneyBee训练的VLM在不同模型规模上均优于最优模型。例如，一个3B参数的HoneyBee训练VLM在MathVerse上分别比最优模型和基础模型高出7.8%和24.8%。此外，我们提出了一种测试时扩展策略，在不牺牲准确率的情况下将解码成本降低73%。总体而言，本工作为视觉语言推理数据集管理研究提供了改进策略。
英文原文摘要 1558 字符
Recent advances in vision-language models (VLMs) have made them highly effective at reasoning tasks. However, the principles underlying the construction of performant VL reasoning training datasets remain poorly understood. In this work, we introduce several data curation approaches and study their impacts on VL reasoning capabilities by carefully controlling training and evaluation setups. We analyze the effects of context (image and question pair) sources, implement targeted data interventions, and explore scaling up images, questions, and chain-of-thought (CoT) solutions. Our findings reveal that (a) context source strategies significantly affect VLM performance, (b) interventions such as auxiliary signals from image captions and the inclusion of text-only reasoning yield substantial gains, and (c) scaling all data dimensions (e.g., unique questions per image and unique CoTs per image-question pair) consistently improves reasoning capability. Motivated by these insights, we introduce HoneyBee, a large-scale, high-quality CoT reasoning dataset with 2.5M examples consisting 350K image-question pairs. VLMs trained with HoneyBee outperform state-of-the-art models across model sizes. For instance, a HoneyBee-trained VLM with 3B parameters outperforms the SOTA model and the base model by 7.8% and 24.8%, respectively, on MathVerse. Furthermore, we propose a test-time scaling strategy that reduces decoding cost by 73% without sacrificing accuracy. Overall, this work presents improved strategies for VL reasoning dataset curation research.