HuskyDoge
  • About
  • Research
  • Blogs
  • Misc
Xiaohongshu

#LLM Pre-training

Research

Towards Looped Models Done Right — Part I: Topology, Input Injection, Recurrent-State Design
Benhao Huang, Chufan Shi, Junlin Chen, Shicheng Wen, Zhengzhong Liu, Eric Xing, Xuezhe Ma
Jul 31st 2026
Institute of Foundation Models
#Loop Models
#LLM Pre-training
#Mixture-of-Experts

Controlled Ouro-to-Huginn ablations isolate loop placement, input injection, and recurrent-state design, then test whether the findings transfer to compute-matched MoE models.

paper
alphaXiv
Blog
Tweet
Part II
Towards Looped Models Done Right — Part I: Topology, Input Injection, Recurrent-State Design