Geo3R: Mitigating Spatial Reasoning Hallucination in Multimodal Large Language Models

2026-07-23

AI Summary

This work addresses the prevalent issue of “spatial reasoning hallucinations” in multimodal large language models—errors that contradict real-world 3D spatial structures—a problem inadequately mitigated by existing approaches. The study is the first to formally define and specifically target this subclass of relational hallucinations. It introduces Geo3R, a plug-and-play, training-free framework that integrates geometric constraints with structured 3D reasoning to effectively resolve hallucinations arising from viewpoint effects, object orientations, and perspective changes. Evaluated across three benchmarks encompassing 18 distinct tasks, Geo3R substantially reduces spatial reasoning hallucinations and outperforms current methods, while demonstrating flexible compatibility with diverse multimodal large language models.

Abstract

Despite remarkable progress in visual understanding, Multimodal Large Language Models (MLLMs) remain prone to hallucinations when reasoning about spatial relationships, often producing judgments that contradict the true 3D structure of the scene. Though several existing works have proposed to mitigate hallucinations, our analysis indicates that they show limited effectiveness in spatial reasoning, as they fail to bridge the fundamental gap between 2D visual representations and 3D spatial reality. Based on this finding, we define hallucinations arising from insufficient spatial structure modeling as spatial reasoning hallucination, a subcategory of relation hallucination that existing mitigation methods fail to address. We further identify three typical scenarios where such hallucinations frequently occur: perspective effects, object orientation, and viewpoint changes. To this end, we propose Geo3R, a training-free, plug-and-play framework that incorporates geometric evidence and structured 3D reasoning to mitigate spatial reasoning hallucination. Experiments on three benchmarks, covering 18 tasks across all three scenarios, show that Geo3R substantially reduces spatial reasoning hallucination across diverse MLLMs without additional training, outperforming existing models and methods.

Problem

• spatial reasoning hallucination
• multimodal large language models
• 3D spatial structure
• relation hallucination
• visual understanding

Innovation

• spatial reasoning hallucination
• geometric evidence
• 3D reasoning
• training-free framework
• multimodal large language models