Muon optimizer learns more from rare data than Adam

Muon optimizer learns more from rare data than Adam

i need to dig deeper. i think the industry is coalescing on:

- Adam is faster
- Muon is more stable
- now, Muon learns rare data better

i wonder if that’s why K2 has that vibe that it has

arxiv.org/abs/2509.26030

btw both Kimi K2 & GLM >=4.5 were trained with Muon

maybe others, not sure. Adam was classically the preferred optimizer. Now, there’s no clear preference anymore. trade-offs all the way down