AI Paper Atlas

Research map / Optimization / Distributed and compressed

MT-DAO: Multi-Timescale Distributed Adaptive Optimizers with Local Updates

Alex Iacob, Andrej Jovanovic, Mher Safaryan, Meghdad Kurmanji, Lorenzo Sani, Samuel Horváth, William F. Shen, Xinchi Qiu, Nicholas D. Lane

ICLR 2026 · Optimization › Distributed and compressed · approach: Communication complexity theory · 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
increase momentum decay beta1 from 0.99 to 0.995 (ablation)MT-DAO · 16M, perplexity vs sync period Kxperplexity increase over Kx=16: +6.2% -> +3.7% at Kx=1024Improved
use MT-DAO (slow momentum plus gradient) instead of single-momentum Local ADOPTLocal ADOPT, 4 workers, K=32 · SmolLM2 held-out val perplexityvalidation perplexity: closes gap to DDP; wall clock reduced 6-27%Improved
MT-DAO versus single-momentum ADOPT-DDP at 720M720M GPT · SmolLM2 val, 720Mtime and tokens to target perplexity: 24% fewer tokens, 35% less timeImproved
MT-DAO versus QHADOPT-DDP across model scalesMT-DAO · SmolLM2 val, 16M/125M/720Mvalidation perplexity: trails at 16M, matches at 125M, about 8% faster wall-clock at 720MMixed
use outer Nesterov momentum with Local ADOPTLocal ADOPT · SmolLM2 val perplexityvalidation perplexityImproved
add slow momentum (beta1=0.999, omega1=0.98) to Local ADOPT (ablation)Local ADOPT · worker update alignmentcosine similarity: > 0.95 across all four metricsImproved
high-beta momenta with higher weights allow larger learning rates (ablation)MT-DAO-ADOPT · 16M LR tuningtolerable learning rateImproved

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