Sparse computation can scale the parametric capacity of LLMs without proportionally increasing active computation, but pretraining sparse architectures remai…
机构:CMU
来源:arXiv 2610.12599 | AI4Papers 论文推荐平台