长 grounded 思考：大规模合成视觉问题和推理链
Long Grounded Thoughts: Synthesizing Visual Problems and Reasoning Chains at Scale
AI总结：
本文提出了一种大规模合成视觉问题和推理链的框架，通过生成高质量数据集提升多模态推理性能，验证了其在视觉、文本和音频任务中的有效性。
摘要：
尽管取得了快速进展，多模态推理仍然缺乏系统的方法来合成大规模以视觉为中心的数据集，超越视觉数学。我们介绍了一个框架，能够合成涵盖不同复杂程度的视觉为中心的问题，并生成超过100万高质量问题的数据集，包括推理轨迹、偏好数据以及支持SFT、离线和在线RL的指令提示。我们的视觉为中心的合成框架采用两阶段过程，专注于：（1）从现有图像中大规模生成多样化的可验证问题；（2）通过合并更简单的问题创建复杂的组合性视觉问题。值得注意的是，在我们的数据上微调Qwen2.5-VL-7B在评估的视觉为中心基准上优于现有的开放数据基线，我们的最佳配置在Vstar Bench、CV-Bench和MMStar-V上匹配或超越了强大的闭合数据模型如MiMo-VL-7B-RL。值得注意的是，尽管数据完全以视觉为中心，它在文本-only推理（MMLU-Pro，+3.7%）和音频推理（MMAU，+1.32%）上也表现出积极的转移效果。同样，尽管数据中不含具身视觉数据，我们在评估开放性具身QA时观察到显著的提升（NiEH，+8.8%）。最后，我们使用我们的数据对整个VLM post-training pipeline进行了大规模分析（100万+），显示（i）在具有认知行为的高质量数据上进行SFT对于扩展在线RL至关重要；（ii）离线RL可以在不增加计算需求的情况下匹配在线RL的性能；（iii）在高质量数据上进行SFT也提高了跨领域、跨模态的转移能力。
英文摘要：
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.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
