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

• URL: http://arxiv.org/abs/2608.03610v2
• Date: Mon, 10 Aug 2026 11:37:39 GMT
• Title: Language-Specialized Multi-Teacher On-Policy Distillation for Multilingual LLM-Based ASR
• Abstract summary: We propose Language-Specialized Multi-Teacher On-Policy Distillation (LS-MOPD), which decouples language-specific knowledge acquisition from multilingual capability integration.<n> Experiments on benchmarks covering Mandarin, Mandarin subdialects, Cantonese, and English demonstrate that LS-MOPD substantially outperforms RL baselines and surpasses the empirical performance envelope defined by the best-performing teachers.
• Score: 17.90532879806511
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• 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, jointly modeling 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), with their expertise then 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 static and dynamic acoustic-prefix configurations 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 surpasses the empirical performance envelope defined by the best-performing RL teachers on nearly all benchmarks, revealing its potential to generalize beyond all teachers in multilingual ASR.

Related papers

• AlignX: Advancing Multilingual Large Language Models with Multilingual Representation Alignment [46.881574083116085]
  Multilingual large language models (LLMs) possess impressive multilingual understanding and generation capabilities.<n>LLMs' performance and cross-lingual alignment often lag for non-dominant languages.<n>We propose AlignX to bridge the multilingual performance gap, which is a two-stage representation-level framework.
  arXiv  Detail & Related papers  (2025-09-29T06:37:46Z)
• Exploring Polyglot Harmony: On Multilingual Data Allocation for Large Language Models Pretraining [16.590296049892576]
  This paper introduces Climb, a novel framework designed to systematically optimize multilingual data allocation.<n>At its core, Climb introduces a cross-lingual interaction-aware language ratio, explicitly quantifying each language's effective allocation by capturing inter-language dependencies.<n>Extensive experiments confirm that Climb can accurately measure cross-lingual interactions across various multilingual settings.
  arXiv  Detail & Related papers  (2025-09-19T03:34:34Z)
• Efficient Multilingual ASR Finetuning via LoRA Language Experts [59.27778147311189]
  This paper proposes an efficient finetuning framework for customized multilingual ASR via prepared LoRA language experts based on Whisper.<n>Through LoRA expert fusion or knowledge distillation, our approach achieves better recognition performance on target languages than standard fine-tuning methods.<n> Experimental results demonstrate that the proposed models yield approximately 10% and 15% relative performance gains in language-aware and language-agnostic scenarios.
  arXiv  Detail & Related papers  (2025-06-11T07:06:27Z)
• LUSIFER: Language Universal Space Integration for Enhanced Multilingual Embeddings with Large Language Models [89.13128402847943]
  We present LUSIFER, a novel zero-shot approach that adapts LLM-based embedding models for multilingual tasks without requiring multilingual supervision.<n>LUSIFER's architecture combines a multilingual encoder, serving as a language-universal learner, with an LLM-based embedding model optimized for embedding-specific tasks.<n>We introduce a new benchmark encompassing 5 primary embedding tasks, 123 diverse datasets, and coverage across 14 languages.
  arXiv  Detail & Related papers  (2025-01-01T15:43:07Z)
• Lens: Rethinking Multilingual Enhancement for Large Language Models [70.85065197789639]
  We propose Lens, a novel approach to enhance multilingual capabilities in large language models (LLMs)<n>Lens operates on two subspaces: the language-agnostic subspace, where it aligns target languages with the central language to inherit strong semantic representations, and the language-specific subspace, where it separates target and central languages to preserve linguistic specificity.<n>Lens significantly improves multilingual performance while maintaining the model's English proficiency, achieving better results with less computational cost compared to existing post-training approaches.
  arXiv  Detail & Related papers  (2024-10-06T08:51:30Z)
• Getting More from Less: Large Language Models are Good Spontaneous Multilingual Learners [67.85635044939836]
  Large Language Models (LLMs) have shown impressive language capabilities. In this work, we investigate the spontaneous multilingual alignment improvement of LLMs. We find that LLMs instruction-tuned on the question translation data (i.e. without annotated answers) are able to encourage the alignment between English and a wide range of languages.
  arXiv  Detail & Related papers  (2024-05-22T16:46:19Z)
• Investigating Multilingual Instruction-Tuning: Do Polyglot Models Demand for Multilingual Instructions? [42.37657013017192]
  We show that instruction-tuning on parallel instead of monolingual corpora benefits cross-lingual instruction following capabilities by up to 9.9%. We also conduct a human annotation study to understand the alignment between human-based and GPT-4-based evaluation within multilingual chat scenarios.
  arXiv  Detail & Related papers  (2024-02-21T11:07:07Z)
• How Vocabulary Sharing Facilitates Multilingualism in LLaMA? [19.136382859468693]
  Large Language Models (LLMs) often show strong performance on English tasks, while exhibiting limitations on other languages. This study endeavors to examine the multilingual capability of LLMs from the vocabulary sharing perspective.
  arXiv  Detail & Related papers  (2023-11-15T16:13:14Z)
• Multilingual DistilWhisper: Efficient Distillation of Multi-task Speech Models via Language-Specific Experts [14.999359332108767]
  We propose DistilWhisper to bridge the performance gap in ASR for under-represented languages. Our approach involves two key strategies: lightweight modular ASR fine-tuning of whisper-small using language-specific experts, and knowledge distillation from whisper-large-v2. Results demonstrate that our approach is more effective than standard fine-tuning or LoRA adapters.
  arXiv  Detail & Related papers  (2023-11-02T08:37:30Z)
• AMTSS: An Adaptive Multi-Teacher Single-Student Knowledge Distillation Framework For Multilingual Language Inference [27.333905128454546]
  AMTSS is an adaptive multi-teacher single-student distillation framework. We first introduce an adaptive learning strategy and teacher importance weight, which enables a student to effectively learn from max-margin teachers. We present a shared student with different projection layers in support of multiple languages, which contributes to largely reducing development and machine cost.
  arXiv  Detail & Related papers  (2023-05-13T14:42:30Z)
• Not All Languages Are Created Equal in LLMs: Improving Multilingual Capability by Cross-Lingual-Thought Prompting [123.16452714740106]
  Large language models (LLMs) demonstrate impressive multilingual capability, but their performance varies substantially across different languages. We introduce a simple yet effective method, called cross-lingual-thought prompting (XLT) XLT is a generic template prompt that stimulates cross-lingual and logical reasoning skills to enhance task performance across languages.
  arXiv  Detail & Related papers  (2023-05-11T17:44:17Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.