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

arXiv: https://arxiv.org/abs/2608.03610 · HTML: https://mlsanity.com/paper/2608.03610/

## Key facts

- Cited by 1 papers in the map (data through 14 Oct 2026).
- Placed in On-policy and self-distillation for LLMs, one of 12 directions in Knowledge distillation.
- Builds on 20 of its 24 references that are in the map.
- First posted on arXiv 4 Aug 2026.

## Placed in

- Knowledge distillation / [On-policy and self-distillation for LLMs](https://mlsanity.com/distill/on-policy-and-self-distillation-for-llms/)

## Builds on (most cited first, 20 of 20)

- [DAPO: An Open-Source LLM Reinforcement Learning System at Scale](https://mlsanity.com/paper/2503.14476/) (2025)
- [ZeRO: Memory Optimizations Toward Training Trillion Parameter Models](https://mlsanity.com/paper/1910.02054/) (2019)
- [DeepSeek-V4: Towards Highly Efficient Million-Token Context Intelligence](https://mlsanity.com/paper/2606.19348/) (2026)
- [On-Policy Distillation of Language Models: Learning from Self-Generated Mistakes](https://mlsanity.com/paper/2306.13649/) (2023)
- [Qwen3-Omni Technical Report](https://mlsanity.com/paper/2509.17765/) (2025)
- [WenetSpeech: A 10000+ Hours Multi-domain Mandarin Corpus for Speech Recognition](https://mlsanity.com/paper/2110.03370/) (2021)
- [MiMo-V2-Flash Technical Report](https://mlsanity.com/paper/2601.02780/) (2026)
- [Qwen3-ASR Technical Report](https://mlsanity.com/paper/2601.21337/) (2026)
- [Step-Audio 2 Technical Report](https://mlsanity.com/paper/2507.16632/) (2025)
- [MOPD: Multi-Teacher On-Policy Distillation for Capability Integration in LLM Post-Training](https://mlsanity.com/paper/2606.30406/) (2026)
- [Seed-ASR: Understanding Diverse Speech and Contexts with LLM-based Speech Recognition](https://mlsanity.com/paper/2407.04675/) (2024)
- [Nemotron-Cascade 2: Post-Training LLMs with Cascade RL and Multi-Domain On-Policy Distillation](https://mlsanity.com/paper/2603.19220/) (2026)
- [Fun-ASR Technical Report](https://mlsanity.com/paper/2509.12508/) (2025)
- [FireRedASR2S: A State-of-the-Art Industrial-Grade All-in-One Automatic Speech Recognition System](https://mlsanity.com/paper/2603.10420/) (2026)
- [WenetSpeech-Yue: A Large-scale Cantonese Speech Corpus with Multi-dimensional Annotation](https://mlsanity.com/paper/2509.03959/) (2025)
- [Investigating the Impact of Cross-lingual Acoustic-Phonetic Similarities on Multilingual Speech Recognition](https://mlsanity.com/paper/2207.03390/) (2022)
- [MOSA: Mixtures of Simple Adapters Outperform Monolithic Approaches in LLM-based Multilingual ASR](https://mlsanity.com/paper/2508.18998/) (2025)
- [Enhancing Multilingual LLM-based ASR with Mixture of Experts and Dynamic Downsampling](https://mlsanity.com/paper/2606.10439/) (2026)
- [Language-Aware Distillation for Multilingual Instruction-Following Speech LLMs with ASR-Only Supervision](https://mlsanity.com/paper/2603.07025/) (2026)
- [NIM4-ASR: Towards Efficient, Robust, and Customizable Real-Time LLM-Based ASR](https://mlsanity.com/paper/2604.18105/) (2026)

## Built on by (most cited first, 1 of 1)

- [Beyond Teacher Assignment: Domain-Normalized Multi-Teacher On-Policy Distillation](https://mlsanity.com/paper/2609.35347/) (2026)

## Abstract (arXiv)

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.