# Distilling 3D Spatial Reasoning into a Lightweight Vision-Language Model with CoT

Source: [Distilling 3D Spatial Reasoning into a Lightweight Vision-Language Model with CoT](<https://www.sota2.com/research/paper/6a02b447dbc836b87d824927>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-05-13T05:00:12.693Z

Authors: Alaa Asfour, Christopher Indris, Leihan Chen, Tejas Vyas, Guanghui Wang

Published: 2026-05-10T19:38:29.000Z

## Abstract

Large-scale 3D vision-language models (VLMs) like LLaVA-3D offer strong spatial reasoning but are difficult to deploy due to high computational costs. We propose a knowledge distillation framework that transfers spatial reasoning from a 7B teacher to a 2.29B student model. Our approach achieves 8.7x lower inference latency and a 3x reduction in model size while retaining 54-72% of the teacher's performance. The framework utilizes VGGT as the vision encoder and a multi-task distillation pipeline with uncertainty-aware loss weighting. To improve reasoning without chain-of-thought (CoT) data, we introduce "Hidden CoT": learnable latent tokens that serve as an internal scratchpad before answer generation. This is the first use of latent scratchpad reasoning in distilled 3D VLMs. The student model jointly performs spatial description, depth estimation, and object detection. Experiments on ScanNet and 3D-FRONT show strong spatial understanding, reaching 68-72% accuracy on proximity and contact tasks. Our framework enables efficient 3D scene QA on resource-constrained platforms.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [3D Spatial Reasoning on ScanNet and 3D-FRONT (test)](<https://www.sota2.com/research/sota/3d-spatial-reasoning-on-scannet-and-3d-front-test.md>): 3D Spatial Reasoning; ScanNet and 3D-FRONT (test); Spatial Accuracy: 64.

- [Text Generation on ScanNet and 3D-FRONT (test)](<https://www.sota2.com/research/sota/text-generation-on-scannet-and-3d-front-test.md>): Text Generation; ScanNet and 3D-FRONT (test); ROUGE-1: 11.9.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2605.09719>)

[PDF](<https://arxiv.org/pdf/2605.09719.pdf>)