ZGCM-1：一個完全開放、極致高效、面向數學與代理式搜尋的基礎模型
ZGCM-1: A Fully Open and Extremely Efficient Foundation Model for Math and Agentic Search
arXiv
15
Liang-Hsun Huang 許願
Puro-2B：窮實驗室在 5090 美元內、用 RTX 5090 訓出的 Qwen2-1.5B
Puro-2B: Poor Lab's Qwen2-1.5B Trained on RTX 5090 within $5090
arXiv
22
Liang-Hsun Huang 許願
Muon 可以擴展到 LLM 訓練
Muon is Scalable for LLM Training
arXiv
35
Liang-Hsun Huang 許願
論 Qwen3.8-Next 架構設計：評測、效率與訓練穩定性
On the Design of Qwen3.8-Next Architecture: Evaluation, Efficiency, and Training Stability
arXiv
1
Liang-Hsun Huang 許願