Sample-efficient Integration of New Modalities into Large Language Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: İnce, Osman Batur, Martins, André F. T., Mac Aodha, Oisin, Ponti, Edoardo M.
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914023104053248
author İnce, Osman Batur
Martins, André F. T.
Mac Aodha, Oisin
Ponti, Edoardo M.
author_facet İnce, Osman Batur
Martins, André F. T.
Mac Aodha, Oisin
Ponti, Edoardo M.
contents Multimodal foundation models can process several modalities. However, since the space of possible modalities is large and evolving over time, training a model from scratch to encompass all modalities is unfeasible. Moreover, integrating a modality into a pre-existing foundation model currently requires a significant amount of paired data, which is often not available for low-resource modalities. In this paper, we introduce a method for sample-efficient modality integration (SEMI) into Large Language Models (LLMs). To this end, we devise a hypernetwork that can adapt a shared projector -- placed between modality-specific encoders and an LLM -- to any modality. The hypernetwork, trained on high-resource modalities (i.e., text, speech, audio, video), is conditioned on a few samples from any arbitrary modality at inference time to generate a suitable adapter. To increase the diversity of training modalities, we artificially multiply the number of encoders through isometric transformations. We find that SEMI achieves a significant boost in sample efficiency during few-shot integration of new modalities (i.e., satellite images, astronomical images, inertial measurements, and molecules) with encoders of arbitrary embedding dimensionality. For instance, to reach the same accuracy as 32-shot SEMI, training the projector from scratch needs 64$\times$ more data. As a result, SEMI holds promise to extend the modality coverage of foundation models.
format Preprint
id arxiv_https___arxiv_org_abs_2509_04606
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Sample-efficient Integration of New Modalities into Large Language Models
İnce, Osman Batur
Martins, André F. T.
Mac Aodha, Oisin
Ponti, Edoardo M.
Computation and Language
Artificial Intelligence
Computer Vision and Pattern Recognition
Multimodal foundation models can process several modalities. However, since the space of possible modalities is large and evolving over time, training a model from scratch to encompass all modalities is unfeasible. Moreover, integrating a modality into a pre-existing foundation model currently requires a significant amount of paired data, which is often not available for low-resource modalities. In this paper, we introduce a method for sample-efficient modality integration (SEMI) into Large Language Models (LLMs). To this end, we devise a hypernetwork that can adapt a shared projector -- placed between modality-specific encoders and an LLM -- to any modality. The hypernetwork, trained on high-resource modalities (i.e., text, speech, audio, video), is conditioned on a few samples from any arbitrary modality at inference time to generate a suitable adapter. To increase the diversity of training modalities, we artificially multiply the number of encoders through isometric transformations. We find that SEMI achieves a significant boost in sample efficiency during few-shot integration of new modalities (i.e., satellite images, astronomical images, inertial measurements, and molecules) with encoders of arbitrary embedding dimensionality. For instance, to reach the same accuracy as 32-shot SEMI, training the projector from scratch needs 64$\times$ more data. As a result, SEMI holds promise to extend the modality coverage of foundation models.
title Sample-efficient Integration of New Modalities into Large Language Models
topic Computation and Language
Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2509.04606