SpatialQuery: Benchmarking Geometry-Grounded Multi-Instance Spatial Reasoning in Vision-Language Models

Multi-View Relational Distillation for Spatial Reasoning with Vision-Language Models

Vision-language models (VLMs) have achieved strong image and video understanding, yet their visual-spatial representations remain geometrically fragile, leading to failures in spatial reasoning needed for embodied AI, robotics, and autonomous driving. Prior approaches to geometry grounding either fine-tune VLMs on spatial question answering, which can perpetuate spurious visual representations, or fuse features from large geometry-grounded vision models, which substantially increases model size at inference. Knowledge distillation from geometry-grounded vision models offers an alternative, but directly matching multi-view teacher features can disrupt the pretrained alignment between visual and textual representations, degrading object- and language-semantic capabilities. We propose multi-view relational distillation (MVRD), which distills patch-wise cosine similarities across views instead of the teacher features themselves. These relations encode geometric correspondences adequate for spatial understanding, while leaving the student representation underdetermined, allowing it to remain close to its pretrained vision- language space. Across representative VLMs, MVRD improves visual-spatial reasoning, outperforming supervised fine-tuning and feature distillation while approaching feature fusion methods with considerably fewer added parameters and lower latency. We show that MVRD makes visual representations more geometric while retaining language alignment, and generalizes to 3D scene understanding tasks such as object grounding, dense captioning, and question answering.

GeoMind: Explicit Spatial Reasoning via Dual-Reference Geometric Modeling

Instance-enriched semantic maps for Visual Language Navigation

Instance-Enriched Semantic Maps is proposed, a unified framework with three key contributions: instance-level two-and-a-half-dimensional rich information mapping, storage-efficient semantic representation that achieves approximately 96% reduction compared to three-dimensional scene-graph approaches while preserving sufficient spatial information for navigation, and robust query processing via LLM-based target selection.

DM-KG: A Novel Method for Boosting Spatial Cognition of Vision-Language Models in Street View Imagery

As vision-language models (VLMs) are increasingly deployed in geospatial question answering and visual scene understanding, improving their spatial cognition capability on street view imagery for complex logical reasoning has emerged as a key research priority. However, existing VLMs frequently suffer from"spatial semantic hallucinations"when perceiving object locations, distances, and directions in real-world street view scenes. Furthermore, such errors are often recalcitrant to tracing and calibration, posing a critical bottleneck for their practical deployment in geospatial tasks. To address this pressing challenge, this study proposes DM-KG (Direction-Metric Knowledge Graph), a structurally grounded spatial representation framework for street view imagery. By explicitly extracting directional and metric relationships between entities from a single 2D image, this framework enhances the spatial reasoning accuracy of VLMs through a structured knowledge graph. Specifically, we integrate panoptic segmentation with metric depth estimation to robustly compute entity-level 3D spatial coordinates. Subsequently, we encode the clock azimuths and Euclidean distances of entity pairs into a JSON-formatted knowledge graph, which is injected into the VLM as an explicit geometric prior to guide spatial reasoning. Experimental results on public spatial question-answering (QA) benchmarks demonstrate that DM-KG reduces the mean absolute error (MAE) in distance estimation by 31.1% and the mean angular error in direction judgment by 65.8%, while simultaneously maintaining a high QA success rate. By establishing a complete, augmented reasoning pipeline, this research significantly improves the spatial cognitive capabilities of VLMs in street view scenarios, thereby providing a flexible, generalized, and interpretable framework for geographic visual question answering (GeoVQA) in open environments.

SpaR3D-MoE: Adaptive 3D Spatial Reasoning from Sparse Views Meets Geometry-Inductive Mixture-of-Experts

Recent Multimodal Large Language Models (MLLMs) struggle to bridge the representational gap between 2D semantic understanding and 3D spatial geometry. Existing 3D-aware models either rely on costly 3D-specific data or utilize RGB-only inputs with heuristic sampling and monolithic, shallow fusion, which respectively disrupt essential spatiotemporal connectivity and induce modality contention across diverse spatial tasks. To overcome these bottlenecks, we introduce SpaR3D-MoE, an end-to-end framework that enables adaptive spatial reasoning by equipping MLLMs with geometry-aware capabilities from only sparse RGB inputs. First, we propose an adaptive spatiotemporal manifold sampling mechanism that constructs a geometry-aware spatiotemporal graph to extract informative keyframes, effectively mitigating sequence redundancy while preserving the scene's topological connectivity. Second, we introduce the heterogeneous geometry-inductive Mixture-of-Experts driven by an instruction-pose aware router, which adaptively routes multimodal tokens to specialized experts, resolving the cross-modal contention inherent in monolithic fusion. Extensive experiments on VSI-Bench, ScanQA, and SQA3D demonstrate that our method achieves state-of-the-art performance. Notably, SpaR3D-MoE achieves the highest average score of 63.5 on VSI-Bench, outperforming the strongest baseline by 7.8 absolute points, alongside relative improvements of 35.4% and 51.4% in Route Plan and Relative Direction tasks, respectively.

3D vision-language question answering with explicit scene graphs and local topology priors

The development of 3D large multimodal models (3D-LMMs) has advanced research on 3D visual question answering. Yet most existing methods rely on implicit feature mapping, where point clouds or scene features are directly projected into the latent space of a language model, without explicitly modeling local spatial structure. In complex indoor environments, this design can lead to spatial judgments that are inconsistent with the actual physical layout during 3D visual question answering, a phenomenon referred to as spatial hallucination. To address this issue, this paper proposes TA-LMM, a 3D visual question answering method built on explicit scene graphs and local topological priors. The method begins by parsing a raw 3D scene into an explicit spatial-semantic scene graph and extracting instance-level representations that encode both semantic features and geometric location information. It then constructs a local physical neighborhood around the target object, serializes neighboring objects together with their distance information into structured priors, and injects them into the multimodal reasoning sequence as conditional context. Under the current experimental setting on the Replica dataset, the results show that the proposed method achieves strong performance on spatial-relation question answering while also alleviating spatial hallucination to a certain extent. These findings suggest that explicit local topological priors can improve scene consistency in 3D visual question answering.