🤖 AI Summary
Existing 3D Gaussian splatting methods struggle to model implicit intentions, complex spatial constraints, and commonsense reasoning, limiting their applicability in embodied interaction scenarios. This work proposes CausalSplat, a novel framework that introduces, for the first time, a comprehensive reasoning task within 3D Gaussian segmentation. By integrating vision-language models with 3D scene graphs, CausalSplat decouples explicit structural perception from implicit logical reasoning to achieve hierarchical holistic understanding. We establish Causal-LERF and Causal-ScanNet, new benchmark datasets encompassing commonsense, spatial, functional, and counterfactual reasoning, on which our method achieves state-of-the-art performance. Furthermore, CausalSplat demonstrates strong generalization capabilities on standard referring and open-vocabulary 3D segmentation tasks.
📝 Abstract
While 3D Gaussian Splatting (3DGS) has advanced open vocabulary scene understanding, existing methods remain confined to explicit queries. They struggle to interpret implicit intents, complex spatial constraints, and commonsense reasoning required for practical embodied interactions. To address this gap, we introduce the task of reasoning 3D Gaussian segmentation and construct two benchmarks, Causal-LERF and Causal-ScanNet. These benchmarks systematically evaluate commonsense, spatial, affordance, and counterfactual reasoning. Evaluations reveal that current state of the art methods perform poorly on these reasoning challenges. Therefore, we propose CausalSplat, a framework that integrates vision-language models with 3D scene graphs to disentangle explicit structural perception from implicit logical inference. Extensive experiments demonstrate that CausalSplat achieves state of the art performance on our reasoning benchmarks while showing strong generalizability on standard referring and open vocabulary 3D segmentation tasks. Project Page: https://jiayuding031020.github.io/CausalSplat