Skip to main navigation Skip to search Skip to main content

Self-Powered LLM Modality Expansion for Large Speech-Text Models

  • Tengfei Yu
  • , Xuebo Liu*
  • , Zhiyi Hou
  • , Liang Ding
  • , Dacheng Tao
  • , Min Zhang
  • *Corresponding author for this work
  • Harbin Institute of Technology Shenzhen
  • Faculty of Computing, Harbin Institute of Technology
  • The University of Sydney
  • Nanyang Technological University

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Large language models (LLMs) exhibit remarkable performance across diverse tasks, indicating their potential for expansion into large speech-text models (LSMs) by integrating speech capabilities. Although unified speech-text pre-training and multimodal data instruction-tuning offer considerable benefits, these methods generally entail significant resource demands and tend to overfit specific tasks. This study aims to refine the use of speech datasets for LSM training by addressing the limitations of vanilla instruction tuning. We explore the instruction-following dynamics within LSMs, identifying a critical issue termed speech anchor bias-a tendency for LSMs to over-rely on speech inputs, mistakenly interpreting the entire speech modality as directives, thereby neglecting textual instructions. To counteract this bias, we introduce a self-powered LSM that leverages augmented automatic speech recognition data generated by the model itself for more effective instruction tuning. Our experiments across a range of speech-based tasks demonstrate that self-powered LSM mitigates speech anchor bias and improves the fusion of speech and text modalities in LSMs. Data, code and scripts are freely available at https://github.com/ytf-philp/Self-powered-LSM.

Original languageEnglish
Title of host publicationEMNLP 2024 - 2024 Conference on Empirical Methods in Natural Language Processing, Proceedings of the Conference
EditorsYaser Al-Onaizan, Mohit Bansal, Yun-Nung Chen
PublisherAssociation for Computational Linguistics (ACL)
Pages12401-12417
Number of pages17
ISBN (Electronic)9798891761643
DOIs
StatePublished - 2024
Externally publishedYes
Event2024 Conference on Empirical Methods in Natural Language Processing, EMNLP 2024 - Hybrid, Miami, United States
Duration: 12 Nov 202416 Nov 2024

Publication series

NameEMNLP 2024 - 2024 Conference on Empirical Methods in Natural Language Processing, Proceedings of the Conference

Conference

Conference2024 Conference on Empirical Methods in Natural Language Processing, EMNLP 2024
Country/TerritoryUnited States
CityHybrid, Miami
Period12/11/2416/11/24

Fingerprint

Dive into the research topics of 'Self-Powered LLM Modality Expansion for Large Speech-Text Models'. Together they form a unique fingerprint.

Cite this