Accelerating Masked Diffusion Large Language Models: A Survey of Efficient Inference Techniques

arXiv:2607.12829v1 Announce Type: cross Abstract: Diffusion large language models (dLLMs) offer a theoretical advantage in parallel generation over standard autoregressive models. However, parallel generation alone does not guarantee practical speedups. Realizing this efficiency requires specialized inference mechanisms, such as diffusion-aware caching and reuse. Consequently, as inference efficiency becomes a prerequisite for practical deployment, recent research has actively explored acceleration techniques across algorithms, architectures, and systems. However, rigorous comparisons remain difficult, as end-to-end latency stems from intricate trade-offs between algorithmic, architectural, and system-level factors that are often conflated in existing benchmarks. In this survey, we introduce a unified latency decomposition framework for dLLMs to disentangle these factors and analyze their impact on inference speed in real deployments. Guided by this framework, we categorize acceleration techniques along three axes covering algorithmic innovations, architectural and system optimizations, and inference-time scaling. Finally, we provide guidelines for reproducible benchmarking and highlight open challenges for realizing the full potential of parallel generation.

Related

• DyLLM: Efficient Diffusion LLM Inference via Saliency-based Token Selection and Partial Attention
• Fast-dLLM++: Frechet Profile Decoding for Faster Diffusion LLM Inference
• DARE: Diffusion Language Model Activation Reuse for Efficient Inference
• dLLM-Cache: Accelerating Diffusion Large Language Models with Adaptive Caching
• Streaming-dLLM: Accelerating Diffusion LLMs via Suffix Pruning and Dynamic Decoding

Source: arXiv cs.AI | 2026-07-15