Signals 4 · free daily AI digest

Mira: Memory-Efficient MoE Inference Using Adaptive Caching and Predictive Expert Staging

Paper recorded by Signals 4 on 2026-09-29 in cs.LG. Abstract reproduced from arXiv; link to the original below.

Published 2026-09-29 on arXiv · recorded by Signals 4 on 2026-09-30

Category: cs.LG · 机器学习 · first seen 2026-09-30

Abstract

Mixture-of-Experts (MoE) models are a compelling architecture for scaling model capacity, making them especially attractive for deployment on resource-constrained, single-GPU systems. However, this benefit is difficult to realize because expert parameters dominate memory, and token-level routing is dynamic, unpredictable, and skewed. Prior work using offloading and caching remains fundamentally re

Read on arXiv →

#12 most recent of 334 cs.LG papers we have recorded · ↑ newer: Dimensionally consistent surrogate modelling through dimensional analy · ↓ older: PDMD: Projected Distribution Matching Distillation for Video Diffusion
Cite this page: Mira: Memory-Efficient MoE Inference Using Adaptive Caching and Predictive Expert Staging: the #12 most recent of 334 cs.LG papers we have recorded (as of 2026-09-29). Source: Signals 4 (Signals API) — https://data.jiangzhang.ca/signals4/t/papers/mira-memory-efficient-moe-inference-using-adaptive-caching-and-predictive-expert.html
Free to quote with attribution to “Signals 4 (Signals API)”. Machine-readable: papers.json
Related: More cs.LG papers · arXiv signals · All papers · Today in AI
Get 4 AI signals a day by email — free.
Subscribe free → See all plans →
Get 4 AI signals a day by email — free
All models · All repos · By company · Daily editions