Advancing Speech Understanding in Speech-Aware Language Models with GRPO

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Elmakies, Avishai, Aronowitz, Hagai, Shabtay, Nimrod, Schwartz, Eli, Hoory, Ron, Dekel, Avihu
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866911166330044416
author Elmakies, Avishai
Aronowitz, Hagai
Shabtay, Nimrod
Schwartz, Eli
Hoory, Ron
Dekel, Avihu
author_facet Elmakies, Avishai
Aronowitz, Hagai
Shabtay, Nimrod
Schwartz, Eli
Hoory, Ron
Dekel, Avihu
contents In this paper, we introduce a Group Relative Policy Optimization (GRPO)-based method for training Speech-Aware Large Language Models (SALLMs) on open-format speech understanding tasks, such as Spoken Question Answering and Automatic Speech Translation. SALLMs have proven highly effective for speech understanding tasks. GRPO has recently gained traction for its efficiency in training LLMs, and prior work has explored its application to SALLMs, primarily in multiple-choice tasks. Building on this, we focus on open-format tasks that better reflect the generative abilities of the models. Our approach leverages GRPO with BLEU as the reward signal to optimize SALLMs, and we demonstrate empirically that it surpasses standard SFT across several key metrics. Finally, we explore the potential of incorporating off-policy samples within GRPO for these tasks, highlighting avenues for further improvement and further research.
format Preprint
id arxiv_https___arxiv_org_abs_2509_16990
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Advancing Speech Understanding in Speech-Aware Language Models with GRPO
Elmakies, Avishai
Aronowitz, Hagai
Shabtay, Nimrod
Schwartz, Eli
Hoory, Ron
Dekel, Avihu
Computation and Language
Artificial Intelligence
Machine Learning
Sound
Audio and Speech Processing
In this paper, we introduce a Group Relative Policy Optimization (GRPO)-based method for training Speech-Aware Large Language Models (SALLMs) on open-format speech understanding tasks, such as Spoken Question Answering and Automatic Speech Translation. SALLMs have proven highly effective for speech understanding tasks. GRPO has recently gained traction for its efficiency in training LLMs, and prior work has explored its application to SALLMs, primarily in multiple-choice tasks. Building on this, we focus on open-format tasks that better reflect the generative abilities of the models. Our approach leverages GRPO with BLEU as the reward signal to optimize SALLMs, and we demonstrate empirically that it surpasses standard SFT across several key metrics. Finally, we explore the potential of incorporating off-policy samples within GRPO for these tasks, highlighting avenues for further improvement and further research.
title Advancing Speech Understanding in Speech-Aware Language Models with GRPO
topic Computation and Language
Artificial Intelligence
Machine Learning
Sound
Audio and Speech Processing
url https://arxiv.org/abs/2509.16990