Cog3DMap: Multi-View Vision-Language Reasoning with 3D Cognitive Maps

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Gwak, Chanyoung, Jeong, Yoonwoo, Jeon, Byungwoo, Lee, Hyunseok, Shin, Jinwoo, Cho, Minsu
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908909577437184
author Gwak, Chanyoung
Jeong, Yoonwoo
Jeon, Byungwoo
Lee, Hyunseok
Shin, Jinwoo
Cho, Minsu
author_facet Gwak, Chanyoung
Jeong, Yoonwoo
Jeon, Byungwoo
Lee, Hyunseok
Shin, Jinwoo
Cho, Minsu
contents Precise spatial understanding from multi-view images remains a fundamental challenge for Multimodal Large Language Models (MLLMs), as their visual representations are predominantly semantic and lack explicit geometric grounding. While existing approaches augment visual tokens with geometric cues from visual geometry models, their MLLM is still required to implicitly infer the underlying 3D structure of the scene from these augmented tokens, limiting its spatial reasoning capability. To address this issue, we introduce Cog3DMap, a framework that recurrently constructs an explicit 3D memory from multi-view images, where each token is grounded in 3D space and possesses both semantic and geometric information. By feeding these tokens into the MLLM, our framework enables direct reasoning over a spatially structured 3D map, achieving state-of-the-art performance on various spatial reasoning benchmarks. Code will be made publicly available.
format Preprint
id arxiv_https___arxiv_org_abs_2603_23023
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Cog3DMap: Multi-View Vision-Language Reasoning with 3D Cognitive Maps
Gwak, Chanyoung
Jeong, Yoonwoo
Jeon, Byungwoo
Lee, Hyunseok
Shin, Jinwoo
Cho, Minsu
Computer Vision and Pattern Recognition
Precise spatial understanding from multi-view images remains a fundamental challenge for Multimodal Large Language Models (MLLMs), as their visual representations are predominantly semantic and lack explicit geometric grounding. While existing approaches augment visual tokens with geometric cues from visual geometry models, their MLLM is still required to implicitly infer the underlying 3D structure of the scene from these augmented tokens, limiting its spatial reasoning capability. To address this issue, we introduce Cog3DMap, a framework that recurrently constructs an explicit 3D memory from multi-view images, where each token is grounded in 3D space and possesses both semantic and geometric information. By feeding these tokens into the MLLM, our framework enables direct reasoning over a spatially structured 3D map, achieving state-of-the-art performance on various spatial reasoning benchmarks. Code will be made publicly available.
title Cog3DMap: Multi-View Vision-Language Reasoning with 3D Cognitive Maps
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2603.23023