@inproceedings{kang-etal-2025-llase,
    title = "{LL}a{SE}-G1: Incentivizing Generalization Capability for {LL}a{MA}-based Speech Enhancement",
    author = "Kang, Boyi  and
      Zhu, Xinfa  and
      Zhang, Zihan  and
      Ye, Zhen  and
      Liu, Mingshuai  and
      Wang, Ziqian  and
      Zhu, Yike  and
      Ma, Guobin  and
      Chen, Jun  and
      Xiao, Longshuai  and
      Weng, Chao  and
      Xue, Wei  and
      Xie, Lei",
    editor = "Che, Wanxiang  and
      Nabende, Joyce  and
      Shutova, Ekaterina  and
      Pilehvar, Mohammad Taher",
    booktitle = "Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingestion-acl-25/2025.acl-long.651/",
    pages = "13292--13305",
    ISBN = "979-8-89176-251-0",
    abstract = "Recent advancements in language models (LMs) have demonstrated strong capabilities in semantic understanding and contextual modeling, which have flourished in generative speech enhancement (SE). However, many LM-based SE approaches primarily focus on semantic information, often neglecting the critical role of acoustic information, which leads to acoustic inconsistency after enhancement and limited generalization across diverse SE tasks. In this paper, we introduce LLaSE-G1, a LLaMA-based language model that incentivizes generalization capabilities for speech enhancement. LLaSE-G1 offers the following key contributions: First, to mitigate acoustic inconsistency, LLaSE-G1 employs continuous representations from WavLM as input and predicts speech tokens from X-Codec2, maximizing acoustic preservation. Second, to promote generalization capability, LLaSE-G1 introduces dual-channel inputs and outputs, unifying multiple SE tasks without requiring task-specific IDs. Third, LLaSE-G1 outperforms prior task-specific discriminative and generative SE models, demonstrating scaling effects at test time and emerging capabilities for unseen SE tasks. Additionally, we release our code and models to support further research in this area."
}