# Polyglot-Lion: Efficient Multilingual ASR for Singapore via Balanced Fine-Tuning of Qwen3-ASR

arXiv: https://arxiv.org/abs/2603.16184 · HTML: https://mlsanity.com/paper/2603.16184/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 24 of its 39 references that are in the map.
- First posted on arXiv 17 Mar 2026.

## Builds on (most cited first, 24 of 24)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [SGDR: Stochastic Gradient Descent with Warm Restarts](https://mlsanity.com/paper/1608.03983/) (2016)
- [Robust Speech Recognition via Large-Scale Weak Supervision](https://mlsanity.com/paper/2212.04356/) (2022)
- [Unsupervised Cross-lingual Representation Learning at Scale](https://mlsanity.com/paper/1911.02116/) (2019)
- [wav2vec 2.0: A Framework for Self-Supervised Learning of Speech Representations](https://mlsanity.com/paper/2006.11477/) (2020)
- [HuBERT: Self-Supervised Speech Representation Learning by Masked Prediction of Hidden Units](https://mlsanity.com/paper/2106.07447/) (2021)
- [Conformer: Convolution-augmented Transformer for Speech Recognition](https://mlsanity.com/paper/2005.08100/) (2020)
- [Common Voice: A Massively-Multilingual Speech Corpus](https://mlsanity.com/paper/1912.06670/) (2019)
- [Qwen-Audio: Advancing Universal Audio Understanding via Unified Large-Scale Audio-Language Models](https://mlsanity.com/paper/2311.07919/) (2023)
- [SALMONN: Towards Generic Hearing Abilities for Large Language Models](https://mlsanity.com/paper/2310.13289/) (2023)
- [FLEURS: Few-shot Learning Evaluation of Universal Representations of Speech](https://mlsanity.com/paper/2205.12446/) (2022)
- [Scaling Speech Technology to 1,000+ Languages](https://mlsanity.com/paper/2305.13516/) (2023)
- [AISHELL-1: An Open-Source Mandarin Speech Corpus and A Speech Recognition Baseline](https://mlsanity.com/paper/1709.05522/) (2017)
- [Massively Multilingual Neural Machine Translation in the Wild: Findings and Challenges](https://mlsanity.com/paper/1907.05019/) (2019)
- [SeaLLMs -- Large Language Models for Southeast Asia](https://mlsanity.com/paper/2312.00738/) (2023)
- [Multilingual Speech Recognition With A Single End-To-End Model](https://mlsanity.com/paper/1711.01694/) (2017)
- [A Survey of Code-switched Speech and Language Processing](https://mlsanity.com/paper/1904.00784/) (2019)
- [Balancing Training for Multilingual Neural Machine Translation](https://mlsanity.com/paper/2004.06748/) (2020)
- [Are Multilingual Models Effective in Code-Switching?](https://mlsanity.com/paper/2103.13309/) (2021)
- [Bytes are All You Need: End-to-End Multilingual Speech Recognition and Synthesis with Bytes](https://mlsanity.com/paper/1811.09021/) (2018)
- [Meta-Transfer Learning for Code-Switched Speech Recognition](https://mlsanity.com/paper/2004.14228/) (2020)
- [Subword Dictionary Learning and Segmentation Techniques for Automatic Speech Recognition in Tamil and Kannada](https://mlsanity.com/paper/2207.13331/) (2022)
- [Knowledge-driven Subword Grammar Modeling for Automatic Speech Recognition in Tamil and Kannada](https://mlsanity.com/paper/2207.13333/) (2022)
- [Towards Improving the Expressiveness of Singing Voice Synthesis with BERT Derived Semantic Information](https://mlsanity.com/paper/2308.16836/) (2023)

## Abstract (arXiv)

We present Polyglot-Lion, a family of compact multilingual automatic speech recognition (ASR) models tailored for the linguistic landscape of Singapore, covering English, Mandarin, Tamil, and Malay. Our models are obtained by fine-tuning Qwen3-ASR-0.6B and Qwen3-ASR-1.7B exclusively on publicly available speech corpora, using a balanced sampling strategy that equalizes the number of training utterances per language and deliberately omits language-tag conditioning so that the model learns to identify languages implicitly from audio. On 12 benchmarks spanning the four target languages, Polyglot-Lion-1.7B achieves an average error rate of 14.85, competitive with MERaLiON-2-10B-ASR (14.32) - a model 6x larger - while incurring a training cost of 81 on a single RTX PRO 6000 GPU compared to 18,862 for the 128-GPU baseline. Inference throughput is approximately 20x faster than MERaLiON at 0.10 s/sample versus 2.02 s/sample. These results demonstrate that linguistically balanced fine-tuning of moderate-scale pretrained models can yield deployment-ready multilingual ASR at a fraction of the cost of larger specialist systems.