Spatial reasoning is a critical capability for intelligent robots, yet current vision-language models (VLMs) still fall short of human-level performance in video-based spatial reasoning. This gap mainly stems from two challenges: a semantic-geometric misalignment that prevents consistent 3D understanding, and the absence of persistent memory to retain 3D representation and understanding over time. To address these limitations, we present VLM2, a Vision-Language Model with persistent Memory for spatial reasoning with a view-consistent, 3D-aware representation purely from 2D video. Specifically, to enhance long-horizon reasoning, we incorporate a dual-memory module, consisting of a working memory that operates as a sliding window to focus on immediate context, and an episodic memory that consolidates and stores critical long-term information. This design enables efficient and long-horizon spatial reasoning with a fixed computational cost. Extensive experiments on multiple benchmarks show that VLM2 achieves state-of-the-art performance among video-only models, significantly advancing the frontier of visual-spatial intelligence.
@article{arxiv.2511.20644,
title = {Vision-Language Memory for Spatial Reasoning},
author = {Zuntao Liu and Yi Du and Taimeng Fu and Shaoshu Su and Cherie Ho and Chen Wang},
journal= {arXiv preprint arXiv:2511.20644},
year = {2025}
}