A latent recurrent language model applies a shared block of layers repeatedly, so one checkpoint can trade inference compute for accuracy by changing the loo…
机构:NVIDIA
来源:arXiv 2610.12582 | AI4Papers 论文推荐平台