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

Quy-Anh Dang, Chris Ngo
3/17/2026
cs.CL

No official code found. Best implementation: knoveleng/polyglot-lion (7 stars, MIT, last commit 2026-03-24).

Abstract

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.

View on arXivView PDF

Code Implementations(1)

knoveleng/polyglot-lionCommunity71%
MIT

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

71Mar 16, 20267 months agoMIT

Similar Papers

LoASR-Bench: Evaluating Large Speech Language Models on Low-Resource Automatic Speech Recognition Across Language Families

88% match2026

Zipper-LoRA: Dynamic Parameter Decoupling for Speech-LLM based Multilingual Speech Recognition

86% match2026

FunAudioLLM: Voice Understanding and Generation Foundation Models for Natural Interaction Between Humans and LLMs

86% match2 repos21.8k2024

DiscoPhon: Benchmarking the Unsupervised Discovery of Phoneme Inventories With Discrete Speech Units

85% match2026

Zero-AVSR: Zero-Shot Audio-Visual Speech Recognition with LLMs by Learning Language-Agnostic Speech Representations

85% match1 repo362025

Learning from Child-Directed Speech in Two-Language Scenarios: A French-English Case Study

84% match2026

Cite this paper

@article{dang2026polyglotlion,
  title  = {Polyglot-Lion: Efficient Multilingual ASR for Singapore via Balanced Fine-Tuning of Qwen3-ASR},
  author = {Quy-Anh Dang and Chris Ngo},
  year   = {2026},
  eprint = {2603.16184},
  archivePrefix = {arXiv},
  url    = {http://arxiv.org/abs/2603.16184v1}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/2603.16184.svg)](https://codeofpaper.com/paper/2603.16184)

Discussion