MoLoRA: Boosting LLM-based End-to-end Speech Translation with Mixture of Low-rank Experts

Hao Zhang; Yaqi Chen; Nianwen Si; XuKui Yang; Wenlin Zhang; Dan Qu

AAAI 2026technical0 citations

MoLoRA: Boosting LLM-based End-to-end Speech Translation with Mixture of Low-rank Experts

Hao Zhang, Yaqi Chen, Nianwen Si, XuKui Yang, Wenlin Zhang, Dan Qu

Abstract

Recently, End-to-End Speech Translation (E2E-ST) methods leveraging large language models (LLMs) have demonstrated strong generalization capabilities and excellent scalability by integrating pre-trained speech encoders with LLMs, where Low-Rank Adaptation (LoRA) is commonly used for parameter-efficient fine-tuning to reduce training costs. However, LoRA

BibTeX

@inproceedings{aaai2026_moloraboostingll,
  title = {MoLoRA: Boosting LLM-based End-to-end Speech Translation with Mixture of Low-rank Experts},
  author = {Hao Zhang and Yaqi Chen and Nianwen Si and XuKui Yang and Wenlin Zhang and Dan Qu},
  booktitle = {AAAI 2026},
  year = {2026}
}