Rank-3 factorization, shared-A tied-KV, RMSNorm, tied embed, curriculum learning
The Mini is a bite-sized version of The New York Times' revered daily crossword. While the crossword is a lengthier experience that requires both knowledge and patience to complete, The Mini is an entirely different vibe.
，这一点在旺商聊官方下载中也有详细论述
FT App on Android & iOS
Stories and lessons learned from an impossibly large community modding project.