AI Paper Atlas

Research map / Architectures and efficiency / Quantization

Training Dynamics Impact Post-Training Quantization Robustness

Albert Catalan-Tatjer, Niccolò Ajroldi, Jonas Geiping

ICLR 2026 · Architectures and efficiency › Quantization · approach: Post-training quantization of LLMs · 8 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
raise the peak learning rate during pretraining (ablation)160M transformer on FineWebEdu, GPTQ 3/4-bit · PTQ error vs validation lossquantization errorImproved
average checkpoints along the trajectory (LAWA) instead of learning rate cooldown (ablation)160M transformer, constant LR, 100B tokens · 3-bit validation loss3-bit validation lossImproved
weight averaging versus LR cooldown in full precision (ablation)160M transformer, constant LR · FP validation lossvalidation lossHurt
apply LAWA to OLMo-1B released checkpointsOLMo-1B · validation loss FP and 4-bitvalidation lossImproved
increase weight decay lambda at comparable full-precision loss (ablation)160M transformer, WSD, LR 3e-3 · 4-bit and 3-bit quantization errorquantization errorImproved
use WSD instead of cosine schedule when scaling token budget (ablation)70M and 160M transformers · delta PTQ vs token budgetdelta_PTQImproved
vary training token budget from 10B to 100B under WSD (ablation)160M transformer · 4-bit quantization error after cooldownquantization errorNo clear effect
replace AdamW with AdamC to reduce loss-gradient norm spike (ablation)160M transformer, WSD · 4-bit quantization errorquantization errorNo clear effect
model soup of OLMo2 decay runs versus individual runs (ablation)OLMo2 family · 3-bit quantization errorquantization errorImproved
merge APO model with mid-training checkpoint (0.9/0.1) for final SmolLM3SmolLM3 · 3-bit degradationvalidation loss and accuracy degradationImproved
cosine versus WSD schedule at higher peak LR (ablation)160M transformer · 4-bit quantization errorquantization errorMixed

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