DFlash: Block Diffusion for Flash Speculative Decoding
This addresses the problem of slow inference for users of large language models, offering a significant speedup over existing methods.
The paper tackles the high inference latency of autoregressive large language models by introducing DFlash, a speculative decoding framework that uses a block diffusion model for parallel drafting, achieving over 6x lossless acceleration and up to 2.5x higher speedup than the state-of-the-art method.
Autoregressive large language models (LLMs) deliver strong performance but require inherently sequential decoding, leading to high inference latency and poor GPU utilization. Speculative decoding mitigates this bottleneck by using a fast draft model whose outputs are verified in parallel by the target LLM; however, existing methods still rely on autoregressive drafting, which remains sequential and limits practical speedups. Diffusion LLMs offer a promising alternative by enabling parallel generation, but current diffusion models typically underperform compared with autoregressive models. In this paper, we introduce DFlash, a speculative decoding framework that employs a lightweight block diffusion model for parallel drafting. By generating draft tokens in a single forward pass and conditioning the draft model on context features extracted from the target model, DFlash enables efficient drafting with high-quality outputs and higher acceptance rates. Experiments show that DFlash achieves over 6x lossless acceleration across a range of models and tasks, delivering up to 2.5x higher speedup than the state-of-the-art speculative decoding method EAGLE-3.