Saved in:
Bibliographic Details
Main Authors: Huynh, Ngoc Dung, Bouadjenek, Mohamed Reda, Razzak, Imran, Hacid, Hakim, Aryal, Sunil
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2503.24164
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911040560693248
author Huynh, Ngoc Dung
Bouadjenek, Mohamed Reda
Razzak, Imran
Hacid, Hakim
Aryal, Sunil
author_facet Huynh, Ngoc Dung
Bouadjenek, Mohamed Reda
Razzak, Imran
Hacid, Hakim
Aryal, Sunil
contents Large vision and language models show strong performance in tasks like image captioning, visual question answering, and retrieval. However, challenges remain in integrating speech, text, and vision into a unified model, especially for spoken tasks. Speech generation methods vary (some produce speech directly), others through text (but their impact on quality is unclear). Evaluation often relies on automatic speech recognition, which may introduce bias. We propose SVLA, a unified speech vision language model based on a transformer architecture that handles multimodal inputs and outputs. We train it on 38.2 million speech text image examples, including 64.1 hours of synthetic speech. We also introduce Speech VQA Accuracy, a new metric for evaluating spoken responses. SVLA improves multimodal understanding and generation by better combining speech, vision, and language.
format Preprint
id arxiv_https___arxiv_org_abs_2503_24164
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SVLA: A Unified Speech-Vision-Language Assistant with Multimodal Reasoning and Speech Generation
Huynh, Ngoc Dung
Bouadjenek, Mohamed Reda
Razzak, Imran
Hacid, Hakim
Aryal, Sunil
Multimedia
Large vision and language models show strong performance in tasks like image captioning, visual question answering, and retrieval. However, challenges remain in integrating speech, text, and vision into a unified model, especially for spoken tasks. Speech generation methods vary (some produce speech directly), others through text (but their impact on quality is unclear). Evaluation often relies on automatic speech recognition, which may introduce bias. We propose SVLA, a unified speech vision language model based on a transformer architecture that handles multimodal inputs and outputs. We train it on 38.2 million speech text image examples, including 64.1 hours of synthetic speech. We also introduce Speech VQA Accuracy, a new metric for evaluating spoken responses. SVLA improves multimodal understanding and generation by better combining speech, vision, and language.
title SVLA: A Unified Speech-Vision-Language Assistant with Multimodal Reasoning and Speech Generation
topic Multimedia
url https://arxiv.org/abs/2503.24164