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

TL;DR

Polyglot-Lion is a compact, multilingual ASR model tailored for Singapore's four main languages, achieving competitive accuracy with significantly lower training costs and faster inference by balanced fine-tuning of moderate-sized pretrained models.

Contribution

The paper introduces a balanced fine-tuning approach for multilingual ASR that enables high performance with reduced computational resources and no explicit language tagging.

Findings

Achieves 14.85% average error rate on 12 benchmarks.

Training cost is significantly lower at $81 compared to $18,862 for baseline.

Inference is approximately 20 times faster than larger models.

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.