Saved in:
Bibliographic Details
Main Authors: Chen, Tianyu, Fu, Xingcheng, Gao, Yisen, Qian, Haodong, Wei, Yuecen, Yan, Kun, Zhou, Haoyi, Li, Jianxin
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2503.18578
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908377667338240
author Chen, Tianyu
Fu, Xingcheng
Gao, Yisen
Qian, Haodong
Wei, Yuecen
Yan, Kun
Zhou, Haoyi
Li, Jianxin
author_facet Chen, Tianyu
Fu, Xingcheng
Gao, Yisen
Qian, Haodong
Wei, Yuecen
Yan, Kun
Zhou, Haoyi
Li, Jianxin
contents Modern vision-language models (VLMs) develop patch embedding and convolution backbone within vector space, especially Euclidean ones, at the very founding. When expanding VLMs to a galaxy scale for understanding astronomical phenomena, the integration of spherical space for planetary orbits and hyperbolic spaces for black holes raises two formidable challenges. a) The current pre-training model is confined to Euclidean space rather than a comprehensive geometric embedding. b) The predominant architecture lacks suitable backbones for anisotropic physical geometries. In this paper, we introduced Galaxy-Walker, a geometry-aware VLM, for the universe-level vision understanding tasks. We proposed the geometry prompt that generates geometry tokens by random walks across diverse spaces on a multi-scale physical graph, along with a geometry adapter that compresses and reshapes the space anisotropy in a mixture-of-experts manner. Extensive experiments demonstrate the effectiveness of our approach, with Galaxy-Walker achieving state-of-the-art performance in both galaxy property estimation ($R^2$ scores up to $0.91$) and morphology classification tasks (up to $+0.17$ F1 improvement in challenging features), significantly outperforming both domain-specific models and general-purpose VLMs.
format Preprint
id arxiv_https___arxiv_org_abs_2503_18578
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Galaxy Walker: Geometry-aware VLMs For Galaxy-scale Understanding
Chen, Tianyu
Fu, Xingcheng
Gao, Yisen
Qian, Haodong
Wei, Yuecen
Yan, Kun
Zhou, Haoyi
Li, Jianxin
Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
Modern vision-language models (VLMs) develop patch embedding and convolution backbone within vector space, especially Euclidean ones, at the very founding. When expanding VLMs to a galaxy scale for understanding astronomical phenomena, the integration of spherical space for planetary orbits and hyperbolic spaces for black holes raises two formidable challenges. a) The current pre-training model is confined to Euclidean space rather than a comprehensive geometric embedding. b) The predominant architecture lacks suitable backbones for anisotropic physical geometries. In this paper, we introduced Galaxy-Walker, a geometry-aware VLM, for the universe-level vision understanding tasks. We proposed the geometry prompt that generates geometry tokens by random walks across diverse spaces on a multi-scale physical graph, along with a geometry adapter that compresses and reshapes the space anisotropy in a mixture-of-experts manner. Extensive experiments demonstrate the effectiveness of our approach, with Galaxy-Walker achieving state-of-the-art performance in both galaxy property estimation ($R^2$ scores up to $0.91$) and morphology classification tasks (up to $+0.17$ F1 improvement in challenging features), significantly outperforming both domain-specific models and general-purpose VLMs.
title Galaxy Walker: Geometry-aware VLMs For Galaxy-scale Understanding
topic Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2503.18578