Authors
Adila AuliaAhmed DabeerAhn JeongmiAntonyrex SajebanChan Hok Teng AdwinCheng Zi Yi NicholasChoa Hsueh Mei EstherHeng JonathanJann Railey Estrada MontalanLee Chwan RenLeong Wai YiLeong Wei QiLiew RachelLimkonchotiwat PeeratMuhammad Ridzuan Bin MokhtarNagarajan KarthikNg Boon Cheong RaymondNgee Chia TaiNgui Jian GangNguyen Thanh NganOng Tat-Wee DavidPereira MarkPhang Shi Wei BenjaminPoon JosephRengarajan HamsawardhiniSusanto YosephineSutaveephamochanon AnochaTan Choon MengTan Chor Phin EvelynTan Le Min SherylTan Siao Wei JessicaTan YixianTasawong PanuthepTee Jun YunTeng Kok Wai WalterTeo Eng Sipp LeslieTjhi WilliamTuchinda PumeWu DonghangYong XianbinZhang Zhou
Abstract
We introduce Nemotron-SEA-LION-v4.8, a family of Southeast Asian Languages In One Network (SEA-LION) models built upon NVIDIA Nemotron 3. The family includes 30B-A3B and 120B-A12B models, with both continued-pretrained base checkpoints and post-trained variants. We adapt the models using Southeast Asian, reasoning, code, and multilingual parallel data, followed by post-training with supervised fine-tuning and online on-policy distillation. On SEA-HELM, the 30B-A3B model improves the overall SEA score from 46.06 to 51.57, while the 120B-A12B model improves from 49.30 to 63.44. Across seven Southeast Asian languages, we observe broad capability gains with the 120B-A12B model showing broader and more consistent improvements across tasks.
Paper Information
- arXiv ID:
- 2609.18310v3
- Published: