Building capable language models for India requires scalable multilingual data and robust evaluation. We introduce Nemotron-Pretraining-Indic-v1, a synthetic…
机构:NVIDIA
来源:arXiv 2610.26006 | AI4Papers 论文推荐平台