构建百万级视觉推理题库，支持多模态模型训练与分析。
Long Grounded Thoughts: Synthesizing Visual Problems and Reasoning Chains at Scale
- 两阶段生成：先批量生成可验证的视觉问题，再组合成复杂推理题。
- 在100万+问题上训练的模型超越开源基线，媲美闭源强模型。
- 数据可跨模态迁移，提升文本、音频及具身问答性能。
尽管进展迅速，多模态推理仍缺乏系统性方法来构建大规模以视觉为中心的数据集，超越视觉数学范畴。我们提出一个框架，能够规模化合成涵盖多样复杂度的视觉中心型问题，并构建了包含超过100万条高质量问题的数据集，包含推理链、偏好数据和指令提示，支持SFT、离线与在线强化学习。该视觉中心合成框架采用两阶段流程：（1）从现有图像中大规模生成多样化可验证问题；（2）通过合并简单问题构建复杂组合型视觉问题。令人瞩目的是，基于该数据微调Qwen2.5-VL-7B在多个评估基准上表现优于现有开源数据基线，最佳配置在Vstar Bench、CV-Bench和MMStar-V上达到或超越强闭源模型如MiMo-VL-7B-RL。值得注意的是，尽管数据完全为视觉中心，其在纯文本推理（MMLU-Pro，+3.7%）和音频推理（MMAU，+1.32%）上均有正向迁移效果；即便不含具身视觉数据，在开放性具身问答任务（NiEH，+8.8%）中也取得显著提升。最后，我们利用该数据对视觉语言模型后训练全流程进行大规模（100万+）分析，发现：（i）高质量数据上的认知行为推理链监督微调是扩展在线强化学习的关键；（ii）离线强化学习可达到与在线强化学习相当的性能，同时降低计算需求；（iii）高质量数据微调有助于提升跨模态、跨域泛化能力。
原文摘要 · Abstract (English)
Despite rapid progress, multimodal reasoning still lacks a systematic approach to synthesize large-scale vision-centric datasets beyond visual math. We introduce a framework able to synthesize vision-centric problems spanning diverse levels of complexity, and the resulting dataset with over 1M high-quality problems including: reasoning traces, preference data, and instruction prompts supporting SFT, offline and online RL. Our vision-centric synthesis framework uses a two-stage process focusing on: (1) generating diverse verifiable questions from existing images at scale, and (2) creating complex compositional visual problems by merging simpler questions. Remarkably, finetuning Qwen2.5-VL-7B on our data outperforms existing open-data baselines across evaluated vision-centric benchmarks, and our best configurations match or surpass strong closed-data models such as MiMo-VL-7B-RL on Vstar Bench, CV-Bench and MMStar-V. Notably, despite being entirely vision-centric, our data transfers positively to text-only reasoning (MMLU-Pro, +3.7%) and audio reasoning (MMAU, +1.32%), demonstrating its effectiveness. Similarly, despite containing no embodied visual data, we observe notable gains (NiEH, +8.8%) when evaluating open-ended embodied QA. Lastly, we use our data to comprehensively analyze at scale (1M+) the entire VLM post-training pipeline showing that (i) SFT on high-quality data with cognitive behaviors on reasoning traces is essential to scale online RL, (ii) offline RL could match online RL's performance while disaggregating compute demands, and, (iii) SFT on high quality data also improve out-of-domain, cross-modality transfer.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。