MODELPAPER

MERaLiON AudioLLM

Data, evaluation, and model training for Singapore’s audio-language model.

From speech data to an audio-language model.

How can audio-language models understand speech in Singapore’s multilingual context?

My work

Led data preparation and evaluation, and co-led model training within a six-person team at A*STAR I²R.

Outcome

MERaLiON AudioLLM, presented at ACL 2025 System Demonstrations, as part of Singapore’s National Multimodal LLM Programme.

Context and my contribution

My role spanned data preparation, evaluation, and co-leading training with the MERaLiON team. Alongside this model work, I led AudioBench, covering eight tasks and 26 datasets, and the curation and release of Singlish speech data. The paper describes the model; my experience page details my responsibilities.