mlsanityData to 14 Oct 2026

Dion3: Full-Stack Orthogonal Updates

arXiv 2608.11612 ↗First posted 12 Aug 2026cs.LG cs.AIcited by 4

Not on the map. Why?

Abstract (arXiv)

The Muon optimizer incurs a significant overhead cost due to its cubic-time Newton-Schulz orthogonalization step. When weights are sharded, communication overhead compounds this computational cost, eroding the benefits of Muon in many settings. We present Dion3, a revision of Muon that targets this overhead at every level of the stack. Our Gram Newton-Schulz algorithm reduces the FLOP cost of orthogonalization, our CuteDSL kernels accelerate it by exploiting symmetry, and our megabatching strategy reduces communication overhead. Moreover, we propose a simple change to the update rule that cuts costs even further: selecting only a fraction of the momentum matrix's rows to orthogonalize at each step. This update rule improves on Dion (another "compressed" version of Muon), in both speed and performance. Overall, Dion3 matches or improves on the loss achieved by Muon but reduces optimizer step time by up to 6x. Dion3 is available via the dion package (https://github.com/microsoft/dion) as a drop-in replacement for Muon.

Key facts. Cited by 4 papers in the map (data through 14 Oct 2026). Builds on 14 of its 33 references that are in the map. First posted on arXiv 12 Aug 2026.

Builds on

14 of 33 references.

References · 14

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2023PyTorch FSDP: Experiences on Scaling Fully Sharded Data Parallelcited by 660
  3. 2026GLM-5: from Vibe Coding to Agentic Engineeringcited by 587
  4. 2025Muon is Scalable for LLM Trainingcited by 428
  5. 2025The Polar Express: Optimal Matrix Sign Methods and Their Application to the Muon Algorithmcited by 110
  6. 2025NorMuon: Making Muon more efficient and scalablecited by 84
  7. 2025Dion: Distributed Orthonormalized Updatescited by 68
  8. 2025Nemotron-CLIMB: CLustering-based Iterative Data Mixture Bootstrapping for Language Model Pre-trainingcited by 66
  9. 2025MuonBP: Faster Muon via Block-Periodic Orthogonalizationcited by 35
  10. 2025SonicMoE: Accelerating MoE with IO and Tile-aware Optimizationscited by 21
  11. 2026On Surprising Effectiveness of Masking Updates in Adaptive Optimizerscited by 7
  12. 2025Motif 2 12.7B technical reportcited by 4
2 more
  1. 2026Hierarchical Muon: Tiled Newton-Schulz Updates for Efficient Muon Optimizationcited by 2
  2. 2026Laguna M.1/XS.2 Technical Reportcited by 1

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2026: 4.

4 papers.