AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Scaling and Transferability of Annealing Strategies in Large Language Model Training

Siqi Wang, Zhengyu Chen, Teng Xiao, Zheqi Lv, Jinluan Yang, Xunliang Cai, Jingang Wang, Xiaomeng Li

AAAI 2026 · Architectures and efficiency › RNNs and normalization · approach: Neural scaling laws · 1 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 Adam-style momentum term instead of multiplicative CMMT (lambda=0.999) (ablation)Forward-Momentum scaling law · Fit across batch sizes, Dense 1BMAPE (%): 2.151 -> 0.674Improved
use Adam-style momentum term instead of multiplicative CMMT (lambda=0.99) (ablation)Forward-Momentum scaling law · Fit across batch sizes, Dense 1BMAPE (%): 0.681 -> 0.674No clear effect
use Adam-style momentum instead of CMMT (lambda=0.999) (ablation)Forward-Momentum scaling law · Fit across model sizes, Dense LRmax 5e-5 batch 512MAPE (%): 0.943 -> 0.579Improved
use Adam-style momentum instead of CMMT (lambda=0.99) (ablation)Forward-Momentum scaling law · Fit across model sizes, MoE LRmax 2e-4MAPE (%): 0.750 -> 0.957Hurt
use Adam-style momentum instead of CMMT (lambda=0.999) (ablation)Forward-Momentum scaling law · Fit across model sizes, MoE LRmax 2e-3MAPE (%): 0.487 -> 0.574Hurt
use Adam-style momentum instead of CMMT (lambda=0.999) (ablation)Forward-Momentum scaling law · Cosine prediction from WSD fit, Dense 100MMAPE (%): 2.251 -> 0.232Improved
track loss by training steps instead of tokens when batch size exceeds optimal batch size (ablation)loss curve modelling · Dense 500M and MoE 700Mloss curve similarityImproved
increase total training steps from 30k to 300kWSD scheduler · 1B MoEoptimal annealing ratioMixed
increase model size at fixed LRmaxWSD scheduler · Dense and MoEoptimal annealing ratioNo clear effect
train on one dataset and evaluate annealing ratio on validation setsWSD scheduler · 1B MoEoptimal annealing ratioNo clear effect

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