TESU-LLM: Training Speech-LLMs Without Speech via Unified Encoder Alignment

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Kim, Taesoo, Ko, Jong Hwan
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908397418315776
author Kim, Taesoo
Ko, Jong Hwan
author_facet Kim, Taesoo
Ko, Jong Hwan
contents Recent advances in speech-enabled language models have shown promising results in building intelligent voice assistants. However, most existing approaches rely on large-scale paired speech-text data and extensive computational resources, which pose challenges in terms of scalability and accessibility. In this paper, we present \textbf{TESU-LLM}, a novel framework that enables training speech-capable language models using only text data. Our key insight is to leverage a unified encoder that maps semantically equivalent text and speech inputs to a shared latent space. By aligning the encoder output with the embedding space of a LLM via a lightweight projection network, we enable the model to generalize from text-only supervision to speech-based inference. Despite being trained exclusively on text, TESU-LLM achieves strong performance on various speech-related benchmarks, comparable to baseline methods trained with large-scale multimodal datasets and substantial computational resources. These results highlight the effectiveness and efficiency of our approach, offering a scalable path toward building speech LLMs without speech data.
format Preprint
id arxiv_https___arxiv_org_abs_2506_06343
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle TESU-LLM: Training Speech-LLMs Without Speech via Unified Encoder Alignment
Kim, Taesoo
Ko, Jong Hwan
Computation and Language
Artificial Intelligence
Machine Learning
Sound
Audio and Speech Processing
Recent advances in speech-enabled language models have shown promising results in building intelligent voice assistants. However, most existing approaches rely on large-scale paired speech-text data and extensive computational resources, which pose challenges in terms of scalability and accessibility. In this paper, we present \textbf{TESU-LLM}, a novel framework that enables training speech-capable language models using only text data. Our key insight is to leverage a unified encoder that maps semantically equivalent text and speech inputs to a shared latent space. By aligning the encoder output with the embedding space of a LLM via a lightweight projection network, we enable the model to generalize from text-only supervision to speech-based inference. Despite being trained exclusively on text, TESU-LLM achieves strong performance on various speech-related benchmarks, comparable to baseline methods trained with large-scale multimodal datasets and substantial computational resources. These results highlight the effectiveness and efficiency of our approach, offering a scalable path toward building speech LLMs without speech data.
title TESU-LLM: Training Speech-LLMs Without Speech via Unified Encoder Alignment
topic Computation and Language
Artificial Intelligence
Machine Learning
Sound
Audio and Speech Processing
url https://arxiv.org/abs/2506.06343