Looped reasoning models repeatedly apply a shared set of parameters, enabling more computation without increasing the model size. These models also support i…
机构:MIT
来源:arXiv 2609.37432 | AI4Papers 论文推荐平台