AI Paper Atlas

Research map / Architectures and efficiency / Attention and transformers

Mosaic: Unlocking Over 30$\times$ Context Length for Diffusion LLMs Inference via Global Memory Planning and Dynamic Peak Taming

Liang Zheng, Bowen Shi, Yitao Hu, Jiawei Zhang, Ruofan Li, Guotao Yang, Zhixin Zhao, Zhengchao Wang, Sheng Chen, Wenxin Li, Dezhi Ran, Tao Xie and 1 more

ICML 2026 · Architectures and efficiency › Attention and transformers · approach: Sparse attention for long-context LLMs

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
add global memory manager (ablation)Mosaic baselines · dLLM inference rp=0.5Lmax multiple: 1.41x-3.00xImproved
add mask-only logits kernel (ablation)Mosaic with global manager · dLLM inference rp=0.5Lmax multiple: 3.68x-8.04xImproved
add lazy chunking optimizer (ablation)Mosaic with global manager and mask-only · dLLM inference rp=0.5Lmax multiple: 15.30x-32.34xImproved
use lazy chunking instead of static full chunking (ablation)Mosaic · dLLM inference up to 32klatency: 5.5%-21.5% reductionImproved
use lazy chunking instead of static full chunking (ablation)Mosaic · dLLM inferenceLmax: identical LmaxNo clear effect
bottleneck-driven search instead of brute force (ablation)Mosaic · dLLM inferencesearch latency: 0.28%-0.32% of brute-forceImproved
first-fit planning instead of ILP (ablation)Mosaic · dLLM inferenceplanning latency: 0.1%-4.3% of ILPImproved
use first-fit instead of ILP (ablation)Mosaic · dLLM inferenceoverall latency: 7.1%-36.7% reductionImproved
add Mosaic to KV-cache dLLM frameworksFast-dLLM, dInfer, Elastic-Cache · LLaDA-8B RTX 3090context length: 3.75x, 3.00x, 2.27xImproved
apply Mosaic under multi-GPULLaDA · 2 GPUscontext length: 12.30x (TP2), 14.92x (PP2)Improved

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