Disentangling 3D Modeling from Spatial Reasoning

2026-08-05

AI Summary

This work proposes DiSR, a novel framework that explicitly decouples 3D geometric modeling from symbolic reasoning—addressing the poor interpretability and computational inefficiency inherent in traditional end-to-end 3D vision-language approaches. DiSR leverages off-the-shelf 3D perception models to generate structured geometric evidence and employs LoRA-finetuned large language models to perform spatial reasoning solely on this explicit representation. Notably, the method achieves competitive performance on mainstream spatial reasoning benchmarks without requiring large-scale 3D visual question answering (VQA) training data or complex toolchains. By separating perception from reasoning, DiSR significantly enhances model interpretability, modularity, and computational efficiency.

Abstract

In this work, we explore an alternative paradigm for spatial reasoning by explicitly disentangling 3D perception from reasoning, rather than jointly acquiring implicit 3D perception and reasoning through large-scale training. Our key observation is that modern perception models excel at estimating continuous 3D geometry, whereas large language models (LLMs) are particularly effective at compositional and symbolic reasoning. Motivated by these complementary strengths, we propose the Disentangled Spatial Reasoner (DiSR), a simple yet effective framework that reconstructs the physical world into structured 3D evidence using off-the-shelf expert perception models and fine-tunes an LLM with LoRA to perform reasoning solely over this explicit geometric evidence. Without large-scale 3D VQA training or complex tool-use policies, DiSR achieves competitive performance on popular spatial reasoning benchmarks. Beyond its strong performance, DiSR offers improved interpretability, modularity, and computational efficiency, demonstrating that explicit separation of perception and reasoning is a scalable and effective alternative paradigm to end-to-end modeling for spatial intelligence.

Problem

• spatial reasoning
• 3D perception
• disentanglement
• modularity
• interpretability

Innovation

• disentangled reasoning
• 3D perception
• large language models
• structured geometric evidence
• LoRA