TextME: Bridging Unseen Modalities Through Text Descriptions

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Hong, Soyeon, Kim, Jinchan, You, Jaegook, Choi, Seungtaek, Kwak, Suha, Cho, Hyunsouk
Format: Preprint
Veröffentlicht: 2026
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866911462286426112
author Hong, Soyeon
Kim, Jinchan
You, Jaegook
Choi, Seungtaek
Kwak, Suha
Cho, Hyunsouk
author_facet Hong, Soyeon
Kim, Jinchan
You, Jaegook
Choi, Seungtaek
Kwak, Suha
Cho, Hyunsouk
contents Expanding multimodal representations to novel modalities is constrained by reliance on large-scale paired datasets (e.g., text-image, text-audio, text-3D, text-molecule), which are costly and often infeasible in domains requiring expert annotation such as medical imaging and molecular analysis. We introduce TextME, the first text-only modality expansion framework, to the best of our knowledge, projecting diverse modalities into LLM embedding space as a unified anchor. Our approach exploits the geometric structure of pretrained contrastive encoders to enable zero-shot cross-modal transfer using only text descriptions, without paired supervision. We empirically validate that such consistent modality gaps exist across image, video, audio, 3D, X-ray, and molecular domains, demonstrating that text-only training can preserve substantial performance of pretrained encoders. We further show that our framework enables emergent cross-modal retrieval between modality pairs not explicitly aligned during training (e.g., audio-to-image, 3D-to-image). These results establish text-only training as a practical alternative to paired supervision for modality expansion.
format Preprint
id arxiv_https___arxiv_org_abs_2602_03098
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle TextME: Bridging Unseen Modalities Through Text Descriptions
Hong, Soyeon
Kim, Jinchan
You, Jaegook
Choi, Seungtaek
Kwak, Suha
Cho, Hyunsouk
Machine Learning
Artificial Intelligence
Expanding multimodal representations to novel modalities is constrained by reliance on large-scale paired datasets (e.g., text-image, text-audio, text-3D, text-molecule), which are costly and often infeasible in domains requiring expert annotation such as medical imaging and molecular analysis. We introduce TextME, the first text-only modality expansion framework, to the best of our knowledge, projecting diverse modalities into LLM embedding space as a unified anchor. Our approach exploits the geometric structure of pretrained contrastive encoders to enable zero-shot cross-modal transfer using only text descriptions, without paired supervision. We empirically validate that such consistent modality gaps exist across image, video, audio, 3D, X-ray, and molecular domains, demonstrating that text-only training can preserve substantial performance of pretrained encoders. We further show that our framework enables emergent cross-modal retrieval between modality pairs not explicitly aligned during training (e.g., audio-to-image, 3D-to-image). These results establish text-only training as a practical alternative to paired supervision for modality expansion.
title TextME: Bridging Unseen Modalities Through Text Descriptions
topic Machine Learning
Artificial Intelligence
url https://arxiv.org/abs/2602.03098