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

TL;DR

This paper explores how structured multi-stage prompting and reinforcement learning enhance the spatial reasoning abilities of vision-language models, leading to better accuracy and robustness, especially under out-of-distribution conditions.

Contribution

It introduces SceneGraph CoT prompting and applies Group Relative Policy Optimization to improve spatial reasoning and generalization in vision-language models.

Findings

Structured scene graph prompting improves spatial reasoning accuracy.

Reinforcement learning with GRPO outperforms supervised fine-tuning.

Models trained with GRPO show better robustness to phrasing variations.

Abstract

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