PruneGround: Plug-and-play Spatial Pruning for 3D Visual Grounding

2026-06-30

AI Summary

This work addresses the challenges of ambiguous predictions and high computational costs in 3D visual grounding within complex scenes, which often arise from processing entire scene data. To this end, the authors propose a plug-and-play language-guided spatial pruning framework (LGSP) that dynamically narrows the search region using natural language instructions. LGSP integrates multi-view conditional description reconstruction (MCDR) with a detection-pretrained spatial large language model (LLM-Grounder) to achieve precise alignment between language and 3D geometry. Notably, the method requires no modification to the backbone model and achieves state-of-the-art performance on nine out of ten benchmarks across all ScanRefer settings and the Nr3D/Sr3D datasets, marking the first effective incorporation of language-guided pruning into 3D visual grounding tasks.

Abstract

3D Visual Grounding (3DVG) aims to localize target objects in 3D scenes given natural language descriptions. Existing approaches typically perform reasoning over the entire scene, leading to ambiguous predictions and high computational cost, especially in cluttered environments. We observe that many referential expressions rely on local spatial context and often correspond to restricted spatial regions rather than the full scene. Motivated by this insight, we propose PruneGround, an effective plug-and-play framework for 3DVG built upon three key components. First, we introduce Language-Guided Spatial Pruning (LGSP), which leverages a frozen Vision Language Model (VLM) to identify language-relevant regions, thereby reducing spatial computation and grounding candidates in the narrower search space. Second, we propose MultiView-Conditioned Description Reformulation (MCDR), which decomposes complex expressions into simplified target-anchor relations and augments missing spatial cues through multi-view reasoning. Finally, we propose LLM-Grounder, which repurposes a detection-pretrained spatial LLM into a language-conditioned grounding model by aligning point cloud and linguistic representations within the pruned region. Extensive experiments on the three most popular point cloud benchmarks demonstrate that our method achieves state-of-the-art results on all three ScanRefer settings and on 9 out of 10 Nr3D/Sr3D settings. Code and models are publicly available: https://github.com/leduckhai/PruneGround

Problem

• 3D Visual Grounding
• spatial pruning
• ambiguous predictions
• computational cost
• cluttered environments

Innovation

• Spatial Pruning
• 3D Visual Grounding
• Vision Language Model
• Multi-view Reasoning
• Language-conditioned Grounding