knoveleng/polyglot-lion-0.6b-v1.5-mlx-8bit

<div align="center">

Polyglot-Lion-0.6B: Compact multilingual ASR for Singapore — English, Mandarin, Tamil & Malay

   

</div>

<div align="center"> <img src="https://knoveleng.github.io/polyglot-lion/assets/overall-error-rate.png" alt="Average error rate comparison across models" width="900"/> </div>

CHANGE LOG: This version was retrained on the same dataset without punctuation removal to improve the model’s ability to recognize pauses and sentence boundaries in speech.

About

Polyglot-Lion-0.6B is a compact multilingual automatic speech recognition (ASR) model tailored for the linguistic landscape of Singapore, covering English, Mandarin, Tamil, and Malay. Developed by Quy-Anh Dang and Chris Ngo at Knovel Engineering, the model is presented in the report "Polyglot-Lion: Efficient Multilingual ASR for Singapore via Balanced Fine-Tuning of Qwen3-ASR".

The model is fine-tuned from Qwen3-ASR-0.6B exclusively on publicly available speech corpora. It achieves an average error rate of 16.52 — halving the base model's error (50.68) and outperforming Whisper-large-v3-turbo (33.04) — and 20× faster inference.

• —Parameters: 0.6B
• —Languages: English, Mandarin, Tamil, Malay
• —Training cost: $81 on a single NVIDIA RTX PRO 6000 (48 h)
• —Inference speed: ~0.10 s/sample on RTX PRO 4500

Methodology

Polyglot-Lion employs a two-stage balanced upsampling strategy to handle severe class imbalance across languages and datasets:

1. 1.Stage 1 (Intra-language balancing): Within each language, smaller datasets are replicated and subsampled to match the largest dataset in that language.
2. 2.Stage 2 (Inter-language balancing): Across languages, per-language corpora are balanced so every language contributes equally to the final training set (~969 hours total).

The model deliberately omits language-tag conditioning during training, allowing it to learn to identify languages implicitly from the audio signal, which is critical for deployment-ready multilingual ASR in diverse linguistic environments.

Results

WER (%) for English, Tamil, and Malay; CER (%) for Mandarin. Lower is better. Bold = best overall.

Quick Start

See mlx-audio for inference.

Citation

@misc{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},
    primaryClass={cs.CL},
    url={https://arxiv.org/abs/2603.16184}, 
}