mlsanityData to 7 Oct 2026

PSD: Pushing the Pareto Frontier of Diffusion LLMs via Parallel Speculative Decoding

arXiv 2605.15609 ↗First posted 15 May 2026cs.CLcited by 0

On the map: Efficient LLM inference and serving → Diffusion LLM and MoE inference

Abstract (arXiv)

Diffusion large language models (dLLMs) generate text by iteratively denoising masked token sequences. Although dLLMs can predict all masked positions in parallel within each step, the large number of denoising iterations still makes inference expensive. This cost can be reduced spatially by unmasking multiple tokens per step, or temporally by collapsing multiple denoising steps into one verification call. We propose Parallel Speculative Decoding (PSD), a training-free framework that jointly improves inference along both axes. Using the confidence scores from a single forward pass, PSD selects positions to unmask via a configurable, adaptive unmasking policy and constructs multi-depth speculative drafts without extra model calls. A final batched verification pass then applies hierarchical acceptance, keeping the deepest draft that remains consistent with the updated predictions. Experiments on three dLLMs across reasoning and code generation tasks show that PSD achieves favorable trade-offs between inference efficiency and generation quality, reaching up to 5.5× tokens per forward pass with accuracy comparable to greedy decoding.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Diffusion LLM and MoE inference, one of 12 directions in Efficient LLM inference and serving. Builds on 45 of its 49 references that are in the map. First posted on arXiv 15 May 2026.

Builds on

45 of 49 references.

Closest ancestors · 8

Same direction, newest first

  1. 2026Why Diffusion Language Models Struggle with Truly Parallel (Non-Autoregressive) Decoding?26 Feb
  2. 2025LoPA: Scaling dLLM Inference via Lookahead Parallel Decoding18 Dec
  3. 2025From Bits to Rounds: Parallel Decoding with Exploration for Diffusion Language Models26 Nov
  4. 2025Accelerating Diffusion LLM Inference via Local Determinism Propagation8 Oct
  5. 2025Structuring The Future: Diffusion LLM Speculative Decoding via Calibrated Draft Graphs22 Sep
  6. 2025Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding28 May
  7. 2025Dimple: Discrete Diffusion Multimodal Large Language Model with Parallel Decoding22 May
  8. 2025Diffusion Models are Secretly Exchangeable: Parallelizing DDPMs via Autospeculation6 May

Other references · 37

  1. 2024The Llama 3 Herd of Modelscited by 14,192
  2. 2021Training Verifiers to Solve Math Word Problemscited by 9,743
  3. 2025Qwen3 Technical Reportcited by 8,195
  4. 2021Evaluating Large Language Models Trained on Codecited by 8,053
  5. 2024Qwen2.5 Technical Reportcited by 4,713
  6. 2021Program Synthesis with Large Language Modelscited by 3,509
  7. 2022Fast Inference from Transformers via Speculative Decodingcited by 1,615
  8. 2023Accelerating Large Language Model Decoding with Speculative Samplingcited by 975
  9. 2025Large Language Diffusion Modelscited by 872
  10. 2024Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Headscited by 821
  11. 2024EAGLE: Speculative Sampling Requires Rethinking Feature Uncertaintycited by 564
  12. 2022Efficiently Scaling Transformer Inferencecited by 529
25 more
  1. 2025Dream 7B: Diffusion Large Language Modelscited by 515
  2. 2024EAGLE-2: Faster Inference of Language Models with Dynamic Draft Treescited by 362
  3. 2025EAGLE-3: Scaling up Inference Acceleration of Large Language Models via Training-Time Testcited by 318
  4. 2025LLaDA 1.5: Variance-Reduced Preference Optimization for Large Language Diffusion Modelscited by 273
  5. 2024Unlocking Efficiency in Large Language Model Inference: A Comprehensive Survey of Speculative Decodingcited by 266
  6. 2025DiffuCoder: Understanding and Improving Masked Diffusion Models for Code Generationcited by 184
  7. 2025LLaDA2.0: Scaling Up Diffusion Language Models to 100Bcited by 154
  8. 2023REST: Retrieval-Based Speculative Decodingcited by 148
  9. 2025Fast-dLLM v2: Efficient Block-Diffusion LLMcited by 128
  10. 2020Data Movement Is All You Need: A Case Study on Optimizing Transformerscited by 108
  11. 2025A Survey on Diffusion Language Modelscited by 75
  12. 2025LLaDA-MoE: A Sparse MoE Diffusion Language Modelcited by 54
  13. 2024Turning Trash into Treasure: Accelerating Inference of Large Language Models with Token Recyclingcited by 48
  14. 2024SuffixDecoding: Extreme Speculative Decoding for Emerging AI Applicationscited by 34
  15. 2025dInfer: An Efficient Inference Framework for Diffusion Language Modelscited by 33
  16. 2025Accelerated Diffusion Models via Speculative Samplingcited by 30
  17. 2025Efficient-DLM: From Autoregressive to Diffusion Language Models, and Beyond in Speedcited by 30
  18. 2025Self Speculative Decoding for Diffusion Large Language Modelscited by 29
  19. 2025Speed Always Wins: A Survey on Efficient Architectures for Large Language Modelscited by 16
  20. 2026SpecForge: A Flexible and Efficient Open-Source Training Framework for Speculative Decodingcited by 15
  21. 2025How Efficient Are Diffusion Language Models? A Critical Examination of Efficiency Evaluation Practicescited by 10
  22. 2026Parallelism and Generation Order in Masked Diffusion Language Models: Limits Today, Potential Tomorrowcited by 8
  23. 2026SPEED-Bench: A Unified and Diverse Benchmark for Speculative Decodingcited by 7
  24. 2025Decoding Large Language Diffusion Models with Foreseeing Movementcited by 6
  25. 2025Scaling Up, Speeding Up: A Benchmark of Speculative Decoding for Efficient LLM Test-Time Scalingcited by 5

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026SpecFold: Folding Multi-Branch Redundancy for Faster Speculative Decoding in Diffusion Language Models
  2. 2024Speculative Diffusion Decoding: Accelerating Language Generation through Diffusion
  3. 2026Trajectory-Level Speculative Decoding for Diffusion Language Models
  4. 2026D^2SD: Accelerating Speculative Decoding with Dual Diffusion Draft Models
  5. 2026SimSD: Simple Speculative Decoding in Diffusion Language Models
  6. 2026AdaFlash: Adaptive Speculative Decoding via On-Policy Distilled Diffusion Drafters
  7. 2025DiffuSpec: Unlocking Diffusion Language Models for Speculative Decoding
  8. 2026LibraSpec: Dynamic Diffusion-Based Speculative Decoding via Marginal-Gain-Driven Optimization