CAPruner: Conceptual-Adjacent Scene Graph Pruner for Enhancing 3D Spatial Reasoning of Large Language Models

ACL ARR 2026 January Submission5747 Authors

Keywords: 3D vision-language, spatial reasoning

Abstract: Large language models (LLMs) have recently been applied to 3D vision-language (3D-VL) tasks, in which spatial reasoning is required to identify target objects based on their positions relative to others (i.e., anchors). To facilitate effective scene layout understanding, scene graphs are commonly used to represent such spatial relations. However, reasoning over full graphs incurs high token costs and computational inefficiencies, motivating the use of scene graph pruning. Existing pruning methods predominantly rely on spatial proximity and often remove task-relevant relations, thereby undermining reliable spatial reasoning. To address these limitations, we derive a key requirement for scene graph pruning: preserving the spatial relations that are most relevant to the specific 3D-VL task. Guided by this insight, we propose the Conceptual-Adjacent Scene Graph Pruner (CAPruner). CAPruner integrates fuzzy semantic relevance with spatial proximity to estimate relation importance, enabling the selection of critical relations in a task-specific context. Moreover, to avoid costly relation-level annotations, CAPruner is trained by supervising the aggregated scores of each node's incident edges. Extensive experiments demonstrate that CAPruner effectively preserves relations essential for spatial reasoning, leading to substantial performance improvements of LLMs on 3D-VL tasks.

Paper Type: Long

Research Area: Multimodality and Language Grounding to Vision, Robotics and Beyond

Research Area Keywords: vision question answering,multimodality

Contribution Types: Model analysis & interpretability

Languages Studied: English

Submission Number: 5747