ExpertMuon-Compass: Alignment-Guided Step Sizes for Mixture-of-Experts Training BTT0FX7XMC
| URI | m2://res/BTT0FX7XMC |
|---|---|
| Kind | preprint |
| Creators | Omatharv Bharat Vaidya, Ashwin Vinod, Pedram Akbarian, Aditya Sai Ellendula, Connor T. Jerzak, Nhat Ho |
| Date | 2026-10-02 |
| Subjects | arxiv cs.LG, arxiv cs.CL, arxiv math.OC |
| Scope | border |
Expressions
e1: version, preprint
| Date | 2026-10-02 |
|---|---|
| Ids | arXiv 2610.04140 |
Items in this resource
The mentions of the inventory that name this resource. Each name links to the card of its inventory row.
| Item | Kind | Where | Source |
|---|---|---|---|
| Five Newton-Schulz iterations | statement | proposition G.1 | arxiv |
| Errors of the cosine | statement | lemma G.3 | arxiv |
| The radius map | statement | lemma G.2 | arxiv |
| Perturbation bound for the family factor and radius | statement | theorem 1 | arxiv |
| P08KYP8FNK | statement | theorem | arxiv |
| Bounds on the family and row factors | statement | lemma G.1 | arxiv |
Catalogue records
- arxiv: 2610.04140 (work)