AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

SPARe: Stacked Parallelism with Adaptive Reordering for Fault-Tolerant LLM Pretraining Systems with 100k+ GPUs

Jin Lee, Zhonghao Chen, Xuhang He, Robert Underwood, Bogdan Nicolae, Franck Cappello, Xiaoyi Lu, Sheng Di, Zheng Zhang

ICML 2026 · Architectures and efficiency › RNNs and normalization · approach: Distributed LLM inference systems · 0 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
stack redundant shards with adaptive reordering instead of traditional replicationSPARe+CKPT vs Rep+CKPT · DES 600k H100, N=200time-to-train / T0: 6.07 -> 2.92Improved
same changeSPARe+CKPT vs Rep+CKPT · DES 600k H100, N=600time-to-train / T0: 4.27 -> 2.49Improved
same changeSPARe+CKPT vs Rep+CKPT · DES 600k H100, N=1000time-to-train / T0: 3.88 -> 2.34Improved
add checkpointing on top of data parallelism (CKPT-only vs redundancy schemes)CKPT-only · DES 600k H100 restart-dominantprogressHurt
increase replication degree beyond r=3 (ablation)Rep+CKPT · DES 600k H100time-to-trainHurt
adaptive reordering instead of full r-stack computation (ablation)SPARe · N=600, r=20computation overhead: 20x -> 2.8xImproved
use SPARe at high redundancySPARe+CKPT · DES 600k H100time-to-train and availabilityImproved
use SPARe at low redundancy (e.g. r=2) under Weibull failuresSPARe+CKPT · DES 600k H100time-to-trainHurt

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