paper-with-me

홈 › Papers

Modality Gap-Driven Subspace Alignment Training Paradigm For Multimodal Large Language Models

2026-02-02 · Xiaomin Yu, Yi Xin, Yuhui Zhang, Wenjie Zhang, Chonghan Liu, Hanzhen Zhao, Chen Liu, Xiaoxing Hu, Ziyue Qiao, Hao Tang, Xiaobin Hu, Chengwei Qin, Hui Xiong, Yu Qiao, Shuicheng Yan arxiv

Despite the success of multimodal contrastive learning in aligning visual and linguistic representations, a persistent geometric anomaly, the Modality Gap, remains: embeddings of distinct modalities expressing identical semantics occupy systematically offset regions. Prior approaches to bridge this gap are largely limited by oversimplified isotropic assumptions, hindering their application in large-scale scenarios. In this paper, we address these limitations by precisely characterizing the geometric shape of the modality gap and leveraging it for efficient model scaling. First, we propose the Fixed-frame Modality Gap Theory, which decomposes the modality gap within a frozen reference frame into stable biases and anisotropic residuals. Guided by this precise modeling, we introduce ReAlign, a training-free modality alignment strategy. Utilizing statistics from massive unpaired data, ReAlign aligns text representation into the image representation distribution via a three-step process comprising Anchor, Trace, and Centroid Alignment, thereby explicitly rectifying geometric misalignment. Building on ReAlign, we propose ReVision, a scalable training paradigm for Multimodal Large Language Models~(MLLMs). ReVision integrates ReAlign into the pretraining stage, enabling the model to learn the distribution of visual representations from unpaired text before visual instruction tuning, without the need for large-scale, high-quality image-text pairs. Our framework demonstrates that statistically aligned unpaired data can effectively substitute for expensive image-text pairs, offering a robust path for the efficient scaling of MLLMs.

📄 PDF Abstract BibTeX arXiv:2602.07026

Code (0)

등록된 구현이 없습니다.

Tasks

Contrastive Learning

Similar Papers 제목 키워드 기반

LION: A Clifford Neural Paradigm for Multimodal-Attributed Graph Learning

2026-01-29 · Xunkai Li, Zhengyu Wu, Zekai Chen, Henan Sun 외 arxiv

Recently, the rapid advancement of multimodal domains has driven a data-centric paradigm shift in graph ML, transitioning from text-attributed to multimodal-attributed graphs. This advancement significantly enhances data…

Graph Learning

Decipher the Modality Gap in Multimodal Contrastive Learning: From Convergent Representations to Pairwise Alignment

2025-09-27 · Lingjie Yi, Raphael Douady, Chao Chen arxiv

Multimodal contrastive learning (MCL) aims to embed data from different modalities in a shared embedding space. However, empirical evidence shows that representations from different modalities occupy completely separate …

Contrastive Learning

Selective Vision-Language Subspace Projection for Few-shot CLIP

2024-07-24 · Xingyu Zhu, Beier Zhu, Yi Tan, Shuo Wang 외

Vision-language models such as CLIP are capable of mapping the different modality data into a unified feature space, enabling zero/few-shot inference by measuring the similarity of given images and texts. However, most e…

Few-Shot Learning

When Language Overwrites Vision: Over-Alignment and Geometric Debiasing in Vision-Language Models

2026-05-07 · Harshvardhan Saini, Samyak Jha, Yiming Tang, Dianbo Liu arxiv

Vision-Language Models (VLMs) increasingly power high-stakes applications, from medical imaging to autonomous systems, yet they routinely hallucinate, confidently describing content not present in the input. We investiga…

IsoCLIP: Decomposing CLIP Projectors for Efficient Intra-modal Alignment

2026-03-20 · Simone Magistri, Dipam Goswami, Marco Mistretta, Bartłomiej Twardowski 외 arxiv

Vision-Language Models like CLIP are extensively used for inter-modal tasks which involve both visual and text modalities. However, when the individual modality encoders are applied to inherently intra-modal tasks like i…

Image Retrieval