Language-Specialized Multi-Teacher On-Policy Distillation for Multilingual LLM-Based ASR

📅 2026-08-04
📈 Citations: 0
Influential: 0
📄 PDF
🤖 AI Summary
This work addresses the challenge of cross-lingual optimization conflicts in multilingual large language model-based speech recognition, where joint training struggles to preserve language-specific characteristics. The authors propose a language-specific multi-teacher online policy distillation framework that integrates language routing with token-level knowledge fusion. To decouple language-specialized capabilities from general multilingual modeling, they introduce both static and dynamic acoustic prefix designs. Evaluated on a mixed benchmark comprising Mandarin, Chinese dialects, Cantonese, and English, the proposed method significantly outperforms reinforcement learning baselines and consistently surpasses all monolingual teacher models, demonstrating superior generalization performance.
📝 Abstract
Modern LLM-based ASR systems have established multilingual capability as a standard feature, leveraging large-scale multilingual corpora and LLMs' cross-lingual knowledge to achieve competitive performance across multilingual benchmarks. However, joint modeling of languages with heterogeneous acoustic, phonological, and lexical characteristics inevitably introduces optimization conflicts, undermining language-wise specialization. To address this challenge, we propose Language-Specialized Multi-Teacher On-Policy Distillation (LS-MOPD), which decouples language-specific knowledge acquisition from multilingual capability integration: language-specialized teachers are independently optimized via reinforcement learning (RL), after which their expertise is integrated into a generalist multilingual student through language routing and token-level multi-teacher distillation, thereby reducing direct cross-lingual optimization conflicts. We further explore two acoustic-prefix configurations, static and dynamic, to examine how teacher--student prefix consistency influences the efficacy of on-policy distillation. Experiments on benchmarks covering Mandarin, Mandarin subdialects, Cantonese, and English demonstrate that LS-MOPD substantially outperforms RL baselines and consistently surpasses the empirical performance envelope defined by best-performing RL teachers, revealing its potential to generalize beyond all teachers in multilingual ASR.
Problem

Research questions and friction points this paper is trying to address.

multilingual ASR
optimization conflicts
language specialization
heterogeneous languages
LLM-based speech recognition
Innovation

Methods, ideas, or system contributions that make the work stand out.

multi-teacher distillation
on-policy distillation
language-specialized ASR
reinforcement learning
multilingual speech recognition
🔎 Similar Papers
2023-11-02IEEE International Conference on Acoustics, Speech, and Signal ProcessingCitations: 20
Y
Yuan Xie
Advanced Intelligent Systems Group, NIO, Beijing, China
J
Jiaqi Song
Advanced Intelligent Systems Group, NIO, Beijing, China
X
Xianliang Wang
Advanced Intelligent Systems Group, NIO, Beijing, China
M
Ming Lei
Advanced Intelligent Systems Group, NIO, Beijing, China
J
Jie Gao
Advanced Intelligent Systems Group, NIO, Beijing, China
J
Jie Wu
Advanced Intelligent Systems Group, NIO, Beijing, China