AI Paper Atlas

Research map / Architectures and efficiency / Pruning and efficient tuning

Taming Momentum: Rethinking Optimizer States Through Low-Rank Approximation

Zhengbo Wang, Jian Liang, Ran He, Zilei Wang, Tieniu Tan

ICLR 2026 · Architectures and efficiency › Pruning and efficient tuning · approach: Low-rank compression and merging · 4 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use LoRA-Pre low-rank online momentum instead of Fira projectionAdam low-rank optimizer · C4 pre-training (Llama 350M)perplexity: 17.22 -> 16.36Improved
use LoRA-Pre Adam instead of Fira at 60MAdam low-rank optimizer · C4 pre-training (Llama 60M)perplexity: 31.19 -> 32.57Hurt
use LoRA-Pre Adam instead of full-rank AdamLlama 1B · C4 pre-training (Llama 1B)perplexity: 15.56 -> 13.53Improved
use LoRA-Pre Adam instead of LoRALlama-2-7B · GSM8K (MetaMath100k fine-tune)accuracy: 44.62 -> 57.35Improved
use LoRA-Pre Adam instead of LoRALlama-3.1-8B · MATH-500 (MetaMath100k fine-tune)accuracy: 17.06 -> 17.66No clear effect
use LoRA-Pre Muon instead of LoRA MuonLlama-2-7B · GSM8K (MetaMath100k fine-tune)accuracy: 35.15 -> 47.20Improved
add momentum compression via LoRA-Pre to Muon instead of GaLore (ablation)Muon low-rank · C4 pre-training (Llama 130M)perplexity: 25.16 -> 23.05Improved
use projection-based low-rank Muon (GaLore) vs no momentum (ablation)Muon · C4 pre-training (Llama 60M)perplexity: 32.15 -> 34.39Hurt
increase rank (ablation)LoRA-Pre and GaLore · C4 pre-training (Llama 60M/130M)perplexityImproved
use LoRA-Pre Adam at rank 16 instead of GaLore at rank 128 (ablation)low-rank optimizer · C4 pre-training (Llama 60M)perplexityImproved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy