HTMuon improves training efficiency for large language models

Original: Great work with tianyu！This is one of the earliest paper apply Σᵖ to Muon.

Source: x.com ↗