From Diagnosis to Improvement: Probing Spatio-Physical Reasoning in Vision Language Models

TL;DR

This paper analyzes the limitations of current vision language models in spatio-physical reasoning, diagnosing their shortcomings, and applying fine-tuning and reinforcement learning to improve their capabilities, though generalization remains challenging.

Contribution

It provides a comprehensive diagnostic of VLMs' spatio-physical reasoning and introduces a fine-tuning and reinforcement learning approach to enhance their performance.

Findings

Current models perform inadequately on spatio-physical reasoning.

Fine-tuning and reinforcement learning significantly improve reasoning capabilities.

Generalization to new physics scenarios remains limited.

Abstract

Spatio-physical reasoning, a foundation capability for understanding the real physics world, is a critical step towards building robust world models. While recent vision language models (VLMs) have shown remarkable progress in specialized domains like multimodal mathematics and pure spatial understanding, their capability for spatio-physical reasoning remains largely unexplored. This paper provides a comprehensive diagnostic analysis of mainstream VLMs, revealing that current models perform inadequately on this crucial task. Further detailed analysis shows that this underperformance is largely attributable to biases caused by human-like prior and a lack of deep reasoning. To address these challenges, we apply supervised fine-tuning followed by rule-based reinforcement learning to Qwen2.5-VL-7B, resulting in significant improvements in spatio-physical reasoning capabilities and surpassing leading proprietary models. Nevertheless, despite this success, the model's generalization to new physics scenarios remains limited -- underscoring the pressing need for new approaches in spatio-physical reasoning.