Fine-Grained Preference Optimization Improves Spatial Reasoning in VLMs

Current Vision-Language Models (VLMs) struggle with fine-grained spatial reasoning, particularly when multi-step logic and precise spatial alignment are required. In this work, we introduce SpatialReasoner-R1, a vision-language reasoning model designed to address these limitations. To construct high-quality supervision for spatial reasoning, we design a Multi-Model Monte Carlo Tree Search (M3CTS) method that generates diverse, logically consistent Long Chain-of-Thought (LongCoT) reasoning trajectories. In addition, we propose a fine-grained Direct Preference Optimization (fDPO) method that introduces segment-specific preference granularity for descriptive grounding and logical reasoning, guided by a spatial reward mechanism that evaluates candidate responses based on visual consistency, spatial grounding, and logical coherence. Experimental results demonstrate that fDPO achieves relative performance gains of 4. 1% and 9. 0% over standard DPO on spatial qualitative and quantitative tasks, respectively. SpatialReasoner-R1, trained with fDPO, sets a new SoTA on SpatialRGPT-Bench, outperforming the strongest baseline by 9. 4% in average accuracy, while maintaining competitive performance on general vision-language tasks.

Authors

• Yifan Shen Siebel School of Computing and Data Science, University of Illinois Urbana-Champaign
• Yuanzhe Liu
• Jingyuan Zhu
• Xu Cao
• Xiaofeng Zhang Harbin Institute of Technology(Shenzhen)
• Yixiao He
• Wenming Ye
• James Rehg
• Ismini Lourentzou University of Illinois at Urbana-Champaign

Keywords

No keywords are indexed for this paper.