---
id: 20261020-T0-05
title: "TRACE：用残差激活码本压缩边缘设备间的Transformer通信"
title_en: "TRACE Compresses Transformer Activations to Speed Distributed Edge Inference"
url: https://ai.daily.yangsir.net/daily/20261020-T0-05
issue_date: 2026-10-20
publish_date: 2026-10-19T04:00:00.000Z
category: research
source_name: "arXiv cs.LG (ML)"
source_url: https://arxiv.org/abs/2610.22435
---

# TRACE：用残差激活码本压缩边缘设备间的Transformer通信

新论文针对多台边缘设备协同推理 Transformer 的场景提出 TRACE，用残差激活码本编码来压缩设备间传输的激活值。论文指出，在序列并行分区下，激活值交换主导了推理延迟，而商用边缘带宽有限，通信成为根本瓶颈。现有压缩方案各有取舍，TRACE 试图在压缩率和精度之间取得更好平衡。实际影响是：同一批边缘设备可以支撑更大的模型或更低的响应延迟。

## English Version

**TRACE Compresses Transformer Activations to Speed Distributed Edge Inference**

A new paper introduces TRACE, which uses residual activation codebook encoding to compress activations exchanged between edge devices running distributed Transformer inference. The authors note that under sequence-parallel partitioning, activation exchange dominates inference latency, and commodity edge bandwidth makes communication the fundamental bottleneck. Existing compression schemes each trade off differently; TRACE aims for a better balance between compression ratio and accuracy. The practical effect: the same fleet of edge devices can support larger models or lower response latency.

---

**来源**：[arXiv cs.LG (ML)](https://arxiv.org/abs/2610.22435)

**详情页**：https://ai.daily.yangsir.net/daily/20261020-T0-05

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*