[论文] Where and Why is Grounded Reasoning Useful?

论文概要

研究领域: CV
作者: David Yan, Le Xue, Forrest Huang, Qiuyi Zhang, Dustin Tran, Andrew M. Dai, Yinfei Yang
发布时间: 2026-11-05
arXiv: 2611.08485

中文摘要

接地推理（grounded reasoning）是指模型在思维链推理中输出相关位置的文本坐标，是解决复杂视觉推理问题的常见策略。然而，接地推理的收益和局限性尚未得到充分研究。我们研究接地推理在何处以及为何最有用。为此，我们引入了一种思维链分析方法来测量模型在其思考过程中使用了多少输出的接地信息，并对接地推理训练配方进行了详细的消融研究。我们发现接地推理改进了计数能力，但在广泛的视觉推理和感知基准测试中，相比无坐标接地的基线没有显示出显著优势。通过几个合成任务，我们表明当任务需要不可描述引用（即难以用语言描述的视觉概念）时，接地推理最有用。然而，我们观察到许多现有的视觉推理基准仅使用可描述的文本引用即可解决。为弥补这一差距，我们引入了 BrickVision，这是一个需要不可描述视觉引用才能解决的新基准。

原文摘要

Grounded reasoning, where a model outputs text coordinates of relevant locations in chain-of-thought reasoning, is a common strategy for solving complex visual reasoning problems. However, the benefits and limitations of grounded reasoning are understudied. We study where and why grounded reasoning is most useful. To do so, we introduce a chain-of-thought analysis method to measure how much a model uses its outputted grounding in its thinking trace and perform a detailed ablation study of a grounded reasoning training recipe. We find that grounded reasoning improves counting but shows no significant gain over baselines without coordinate grounding across a wide range of spatial reasoning and perception benchmarks. Using several synthetic tasks, we show that grounded reasoning is most usefu...

自动采集于 2026-11-07

#论文 #arXiv #CV #小凯