GaugeVLM: Structuring Spatial Supervision with Measured Geometric Interventions

2026-09-29

AI Summary

This study addresses the limitations of vision-language models in spatial reasoning, specifically their susceptibility to cross-view contradictions and failure to respond to relational changes. To overcome these issues, we propose GaugeDPO, a method that constructs explicit spatial supervision signals through controlled geometric interventions within 3D scenes. By mapping measurement errors onto preference margins, this approach establishes a joint optimization framework integrating cross-view consistency with intervention constraints. Experimental results demonstrate significant improvements across ten spatial metrics; notably, a 7B-parameter model achieves gains of 15.0 and 18.9 percentage points on distance estimation and spatial relation tasks, respectively. Furthermore, the proposed method successfully generalizes to downstream applications such as autonomous driving, validating its practical efficacy and robustness in real-world scenarios.

Abstract

Vision-language models (VLMs) can contradict themselves across views of the same spatial relation and fail to respond when that relation changes. Addressing these failures requires supervision that captures error magnitude and geometric dependencies across observations, both of which remain implicit in training on individual answers or ordinal preferences. Therefore, we introduce GaugeVLM, which makes this structure explicit through controlled object and camera interventions in explicit 3D scenes, producing linked observations with measured differences between spatial relations and shared truths across views. To translate this structure into learning signals, its core objective, GaugeDPO, converts measured errors into preference margins, directly supervises correct canonical rankings across views, and links intervention-induced answer-odds contrasts to measured relation changes with view-specific scales. Our analysis bounds canonical prediction error and establishes that the cross-view and intervention constraints can be jointly satisfied. Empirically, GaugeVLM improves all 10 established spatial metrics over supervised fine-tuning across three VLM backbones, with the main 7B model gaining 15.0 and 18.9 percentage points on MSMU distance and QSpatial+, respectively. These gains also extend to autonomous driving and embodied reasoning, demonstrating the robust generalization across domains.

Problem

• Vision-Language Models
• Spatial Reasoning
• Cross-view Consistency
• Geometric Dependencies

Innovation

• Vision-Language Models
• Spatial Reasoning
• Geometric Interventions
• GaugeDPO
• 3D Scene Supervision