AI Paper Atlas

Research map / Architectures and efficiency / Quantization

ASTRA: Communication-Efficient Acceleration for Multi-Device Transformer Inference

Xiao Liu, Lijun Zhang, Deepak Ganesan, Hui Guan

ICML 2026 · Architectures and efficiency › Quantization · approach: Ultra-low-bit LLM quantization · 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
use mixed-precision attention with VQ codes for non-local tokens instead of full-precision sequence parallelism (ablation)ASTRA · latency, 4 devices, 1024 tokens, 10 Mbpslatency speedup over SP: up to 171.82xImproved
add Noise-Augmented Vector Quantization (lambda = 1.0) (ablation)ASTRA VQ · validation accuracy under extreme compressionaccuracy: +0.86%Improved
use distributed class tokens instead of a single class token (ablation)ASTRA ViT · CIFAR-100 and ImageNet-1K ablationaccuracy: +0.37% to +7.13%Improved
increase VQ groups from 1 to 32 (ablation)ASTRA · ImageNet-1K, ViT-Basetop-1 accuracy (%): 77.39 -> 80.28Improved
increase number of devices from 2 to 8 (ablation)ASTRA 32 groups · CIFAR-100, ViT-Basetop-1 accuracy (%): 91.86 -> 91.14Hurt
apply ASTRA to zero-shot Wikitext-103 transferGPT2-M ASTRA 32 groups · Wikitext-103 zero-shotperplexity: 43.22 -> 62.29Hurt
combine ASTRA with 4-bit post-training quantization (ablation)ASTRA 32 groups, ViT-Base · ImageNet-1Kaccuracy (%): 80.28 -> 79.78No clear effect
use ASTRA instead of multi-device baselines under low bandwidthLlama-3-8B prefill, 4 devices · 10-100 Mbpsprefill latency speedup over BP Nb=4: 1.13-5.13xImproved
increase device count from 2 to 8 with ASTRA 1 group at 20 Mbps (ablation)ASTRA · latency speedup, 1024 tokensspeedup over single device: 1.72x -> 3.69xImproved

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