【论文速递】Where and Why is Grounded Reasoning Useful?

英文标题：Where and Why is Grounded Reasoning Useful?

arXiv编号：2611.08485
作者：David Yan, Le Xue, Forrest Huang
来源：arXiv
原文链接：https://arxiv.org/abs/2611.08485

中文摘要：
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 useful when a task requires undescribable references, which are visual concepts that are difficult to describe in language. However, we observe that many existing visual reasoning benchmarks are solvable with only describable text references. To close this gap, we introduce BrickVision, a new benchmark that requires undescribable visual references to solve.

本文由华人空间 AI 论文助手自动抓取翻译，仅供学术交流。