pith. sign in

modded-nanogpt: Speedrunning the nanogpt baseline

1 Pith paper cite this work. Polarity classification is still indexing.

1 Pith paper citing it

citation-role summary

background 1

citation-polarity summary

fields

cs.LG 1

years

2026 1

verdicts

UNVERDICTED 1

roles

background 1

polarities

background 1

representative citing papers

OrScale: Orthogonalised Optimization with Layer-Wise Trust-Ratio Scaling

cs.LG · 2026-05-08 · unverdicted · novelty 6.0

OrScale adds a Frobenius-norm trust-ratio layer-wise scaler to Muon’s orthogonalized updates, with per-layer calibration for language models, yielding higher CIFAR-10 accuracy and better language-model pre-training loss than Muon+Moonlight and AdamW.

citing papers explorer

Showing 1 of 1 citing paper.

  • OrScale: Orthogonalised Optimization with Layer-Wise Trust-Ratio Scaling cs.LG · 2026-05-08 · unverdicted · none · ref 6

    OrScale adds a Frobenius-norm trust-ratio layer-wise scaler to Muon’s orthogonalized updates, with per-layer calibration for language models, yielding higher CIFAR-10 accuracy and better language-model pre-training loss than Muon+Moonlight and AdamW.