DualDiffusion: A Speculative Decoding Strategy for Masked Diffusion Models

TL;DR

DualDiffusion introduces a speculative decoding framework for Masked Diffusion Models, combining fast approximate drafting with accurate verification to improve inference speed without sacrificing quality.

Contribution

It proposes a novel decoding strategy that enhances inference efficiency for MDMs by integrating lightweight draft models with verification, surpassing existing speed-quality trade-offs.

Findings

Maintains high accuracy while reducing generation steps.

Achieves better speed-quality trade-offs compared to prior methods.

Demonstrates effectiveness on MMLU and GSM8K benchmarks.

Abstract

Masked Diffusion Models (MDMs) offer a promising alternative to autoregressive language models by enabling parallel token generation and bidirectional context modeling. However, their inference speed is significantly limited by the inability to cache key-value pairs due to bidirectional attention, requiring $O(N^2)$ computations at each generation step. While recent methods like FastDLLM and DkvCache improve inference speed through attention approximations and caching strategies, they achieve speedups at the cost of generation quality. We propose DualDiffusion, a speculative decoding framework for MDMs that combines fast drafter models (using efficient approximations) with slower, more accurate verifier models. By running multiple steps of a lightweight drafter followed by a single verification step, DualDiffusion achieves a superior Pareto frontier between generation steps and accuracy compared to existing approaches. We evaluate our method on MMLU and GSM8K, demonstrating that DualDiffusion maintains high accuracy while reducing the number of generation steps required, effectively pushing the quality-efficiency trade-off curve for masked diffusion language models.