Search papers, labs, and topics across Lattice.
Polyglot-Lion is a family of multilingual ASR models for Singaporean languages, created by fine-tuning Qwen3-ASR-0.6B and 1.7B on public speech corpora. A balanced sampling strategy, omitting explicit language tags, forces the model to implicitly learn language identification. Polyglot-Lion-1.7B achieves competitive accuracy (14.85% average error rate on 12 benchmarks) compared to a 6x larger model (MERaLiON-2-10B-ASR), while reducing training costs by over 200x and inference latency by 20x.
Get competitive multilingual ASR performance with 6x smaller models and 200x less training cost by using balanced fine-tuning and implicit language learning.
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.
