Flash-dLLM: IO-Aware KV Caching and Parallel Decoding for Fast, Memory-Efficient Diffusion LLMs

Paper recorded by Signals 4 on 2026-09-22 in cs.CL. Abstract reproduced from arXiv; link to the original below.

Published 2026-09-22 on arXiv · recorded by Signals 4 on 2026-09-23

Category: cs.CL · 自然语言处理 · first seen 2026-09-23

Abstract

Diffusion Large Language Models (dLLMs) have recently emerged as a promising alternative to autoregressive LLMs by enabling non-autoregressive text generation. However, their practical deployment remains limited by inefficient inference, largely due to the absence of effective Key-Value (KV) caching and scalable parallel decoding mechanisms. Existing acceleration methods typically study KV caching

Read on arXiv →