HoneyBee：面向视觉-语言推理器的数据配方

HoneyBee: Data Recipes for Vision-Language Reasoners

October 14, 2025

作者: Hritik Bansal, Devandra Singh Sachan, Kai-Wei Chang, Aditya Grover, Gargi Ghosh, Wen-tau Yih, Ramakanth Pasunuru

cs.AI

摘要

近期，视觉-语言模型（VLMs）在推理任务中展现出了卓越的性能。然而，构建高效VL推理训练数据集的基本原则仍不甚明了。本研究提出了多种数据整理方法，并通过严格控制训练与评估设置，探讨了这些方法对VL推理能力的影响。我们分析了上下文（图像与问题对）来源的影响，实施了有针对性的数据干预，并探索了图像、问题及思维链（CoT）解决方案的规模化扩展。研究发现：(a) 上下文来源策略显著影响VLM性能，(b) 如图像描述提供的辅助信号及包含纯文本推理等干预措施能带来显著提升，(c) 所有数据维度（如图像对应唯一问题数量及图像-问题对对应唯一CoT数量）的扩展均能持续增强推理能力。基于这些洞见，我们推出了HoneyBee，一个包含250万样本、35万图像-问题对的大规模高质量CoT推理数据集。使用HoneyBee训练的VLMs在不同模型规模下均超越了现有最先进模型。例如，一个拥有30亿参数的HoneyBee训练VLM在MathVerse上的表现分别比SOTA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，在不牺牲准确性的前提下，将解码成本降低了73%。总体而言，本研究为VL推理数据集整理研究提供了改进策略。

English

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.