arxiv
PublishedMay 27, 2026 at 4:00 AM
MuCon: Clipped Muon Updates for LLM Training
Publisher summary· verbatim
arXiv:2605.26459v1 Announce Type: new Abstract: Muon-style optimizers take a matrix-valued momentum or preconditioned update $B = U \operatorname{diag}(\sigma_1,\ldots,\sigma_r) V^\top$ and replace it with its canonical partial polar factor $\operatorname{Pol}(B) = U V^\top$. This maps every nonzero
Stay posted· Newsletter
A 5-min weekly brief — top movers, price watch, story of the week.
Discussion
No replies yet. Be first.
Related coverage
More from ARXIV
arxivExecRetrieval: Measuring the Functional-Correctness Gap in Code-Embedding Retrieval17marxivMasterControl Seventeen Every Time17marxivSpeculative Macro Commit for Faster Tool-Using Agents17marxivFresh Memory, Stale Plans: Dependency-Scoped Validation for Distributed LLM-Agent Memory17mThe Bubble Brief
WEEKLYRead AI insights every Tuesday — top movers, new releases, story of the week.
Originally published on arxiv ↗