Looped language models improve reasoning and knowledge manipulation by applying shared computation repeatedly. Existing systems usually repeat an entire laye…
机构:UC Santa Cruz
来源:arXiv 2608.18230 | AI4Papers 论文推荐平台