arXivShaowen Wang, Ge Zhang, Kairong Luo, Yuhao Wu, Shaofan Liu, Jiaheng Liu, Wenhao Huang, Shen Yan, Jian LiTue, Sep 1, 2026, 7:52 AM PDT
score 16.4
Reusing middle layers twice makes AI models smarter for same cost
Original: SMELT: Scaling Laws for Compute-Matched MoE Looped Transformers
Source: arxiv.org ↗
Writing ELI5 summary…