arXiv 2507.133622025-07-06cs.CVcs.AIcs.CL

通过思维链提示与强化学习增强视觉语言模型的空间推理能力

Enhancing Spatial Reasoning in Vision-Language Models via Chain-of-Thought Prompting and Reinforcement Learning

Courant New York University New York, United States（Courant 新 York 大学新 York，美国）

Binbin Ji, Siddharth Agrawal, Qiance Tang, Yvonne Wu

备注：: 10 pages, 5 figures, submitted to a conference (IEEE formate). Authored by students from the Courant Institute, NYU

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

本文研究视觉语言模型的空间推理能力，发现简单思维链提示可能损害性能，而基于场景图的SceneGraph CoT提示和GRPO强化学习微调能显著提升空间推理准确率与分布外泛化鲁棒性。

摘要：

本研究通过思维链（CoT）提示和强化学习考察视觉语言模型（VLM）的空间推理能力。我们首先评估不同提示策略的影响，发现简单的 CoT 格式——即模型在给出答案前先生成一个推理步骤——不仅无助于提升表现，甚至可能损害模型的原始性能。相比之下，基于场景图的结构化多阶段提示（SceneGraph CoT）显著提高了空间推理准确率。此外，为了提升空间推理能力，我们在 SAT 数据集上使用分组相对策略优化（GRPO）对模型进行微调，并在 CVBench 上评估其性能。与监督微调（SFT）相比，GRPO 在 Pass@1 评估中取得了更高的准确率，并在分布外（OOD）条件下表现出更强的鲁棒性。特别是，我们发现 SFT 会过拟合于表层语言模式，当测试时的措辞发生变化（例如从“closer to”变为“farther from”）时，性能可能下降。相比之下，GRPO 的泛化更加可靠，并能在这种变化下保持稳定性能。我们的研究结果为强化学习和结构化提示如何改善现代 VLM 的空间推理能力与泛化行为提供了见解。所有代码已在以下地址开源：https://github.com/Yvonne511/spatial-vlm-investigator

英文摘要：

This study investigates the spatial reasoning capabilities of vision-language models (VLMs) through Chain-of-Thought (CoT) prompting and reinforcement learning. We begin by evaluating the impact of different prompting strategies and find that simple CoT formats, where the model generates a reasoning step before the answer, not only fail to help, but can even harm the model's original performance. In contrast, structured multi-stage prompting based on scene graphs (SceneGraph CoT) significantly improves spatial reasoning accuracy. Furthermore, to improve spatial reasoning ability, we fine-tune models using Group Relative Policy Optimization (GRPO) on the SAT dataset and evaluate their performance on CVBench. Compared to supervised fine-tuning (SFT), GRPO achieves higher accuracy on Pass@1 evaluations and demonstrates superior robustness under out-of-distribution (OOD) conditions. In particular, we find that SFT overfits to surface-level linguistic patterns and may degrade performance when test-time phrasing changes (e.g., from "closer to" to "farther from"). GRPO, on the other hand, generalizes more reliably and maintains stable performance under such shifts. Our findings provide insights into how reinforcement learning and structured prompting improve the spatial reasoning capabilities and generalization behavior of modern VLMs. All code is open source at: https://github.com/Yvonne511/spatial-vlm-investigator

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。