mlsanityData to 7 Oct 2026

Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding

arXiv 2505.22618 ↗First posted 28 May 2025cs.CLcited by 404

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

Abstract (arXiv)

Diffusion-based large language models (Diffusion LLMs) have shown promise for non-autoregressive text generation with parallel decoding capabilities. However, the practical inference speed of open-sourced Diffusion LLMs often lags behind autoregressive models due to the lack of Key-Value (KV) Cache and quality degradation when decoding multiple tokens simultaneously. To bridge this gap, we introduce a novel block-wise approximate KV Cache mechanism tailored for bidirectional diffusion models, enabling cache reuse with negligible performance drop. Additionally, we identify the root cause of generation quality degradation in parallel decoding as the disruption of token dependencies under the conditional independence assumption. To address this, we propose a confidence-aware parallel decoding strategy that selectively decodes tokens exceeding a confidence threshold, mitigating dependency violations and maintaining generation quality. Experimental results on LLaDA and Dream models across multiple LLM benchmarks demonstrate up to 27.6× throughput improvement with minimal accuracy loss, closing the performance gap with autoregressive models and paving the way for practical deployment of Diffusion LLMs.

Key facts. Cited by 404 papers in the map (data through 7 Oct 2026). Placed in Diffusion LLM and MoE inference, one of 12 directions in Efficient LLM inference and serving. Builds on 29 of its 36 references that are in the map. First posted on arXiv 28 May 2025.

Builds on

29 of 36 references.

Closest ancestors · 1

Same direction, newest first

  1. 2025Dimple: Discrete Diffusion Multimodal Large Language Model with Parallel Decoding22 May

Other references · 28

  1. 2017Attention Is All You Needcited by 59,932
  2. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,393
  3. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,005
  4. 2019BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehensioncited by 5,874
  5. 2021Zero-Shot Text-to-Image Generationcited by 4,276
  6. 2021Structured Denoising Diffusion Models in Discrete State-Spacescited by 1,665
  7. 2025Large Language Diffusion Modelscited by 872
  8. 2024Simple and Effective Masked Diffusion Language Modelscited by 800
  9. 2024Simplified and Generalized Masked Diffusion for Discrete Datacited by 606
  10. 2021Argmax Flows and Multinomial Diffusion: Learning Categorical Distributionscited by 566
  11. 2022A Continuous Time Framework for Discrete Denoising Modelscited by 437
  12. 2024Your Absorbing Discrete Diffusion Secretly Models the Conditional Distributions of Clean Datacited by 371
16 more
  1. 2023Make-An-Audio: Text-To-Audio Generation with Prompt-Enhanced Diffusion Modelscited by 323
  2. 2024Discrete Flow Matchingcited by 301
  3. 2025LLaDA 1.5: Variance-Reduced Preference Optimization for Large Language Diffusion Modelscited by 273
  4. 2024Scaling Diffusion Language Models via Adaptation from Autoregressive Modelscited by 268
  5. 2022DiffusionBERT: Improving Generative Masked Language Models with Diffusion Modelscited by 214
  6. 2024Masked Diffusion Models are Secretly Time-Agnostic Masked Models and Exploit Inaccurate Categorical Samplingcited by 204
  7. 2024Scaling up Masked Diffusion Models on Textcited by 195
  8. 2022Score-based Continuous-time Discrete Diffusion Modelscited by 167
  9. 2025LLaDA-V: Large Language Diffusion Models with Visual Instruction Tuningcited by 145
  10. 2022Concrete Score Matching: Generalized Score Matching for Discrete Datacited by 144
  11. 2024Energy-Based Diffusion Language Models for Text Generationcited by 86
  12. 2024Discrete Copula Diffusioncited by 53
  13. 2023Diffusion Language Models Can Perform Many Tasks with Scaling and Instruction-Finetuningcited by 39
  14. 2023Diffusion-NAT: Self-Prompting Discrete Diffusion for Non-Autoregressive Text Generationcited by 21
  15. 2023DiSK: A Diffusion Model for Structured Knowledgecited by 16
  16. 2025Ideas in Inference-time Scaling can Benefit Generative Pre-training Algorithmscited by 12

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 121; 2026: 283.

404 papers.

Sort:

By area:

  1. 2025DiffuCoder: Understanding and Improving Masked Diffusion Models for Code Generationcited by 184
  2. 2025Fast-dLLM v2: Efficient Block-Diffusion LLMcited by 128
  3. 2025Diffusion LLMs Can Do Faster-Than-AR Inference via Discrete Diffusion Forcingcited by 111
  4. 2025Discrete Diffusion VLA: Bringing Discrete Diffusion to Action Decoding in Vision-Language-Action Policiescited by 90
  5. 2025Lumina-DiMOO: An Omni Diffusion Large Language Model for Multi-Modal Generation and Understandingcited by 83
  6. 2025Dream-Coder 7B: An Open Diffusion Language Model for Codecited by 81
  7. 2025A Survey on Diffusion Language Modelscited by 75
  8. 2025Diffusion Language Models Know the Answer Before Decodingcited by 68
  9. 2025Sparse-dLLM: Accelerating Diffusion LLMs with Dynamic Cache Evictioncited by 64
  10. 2025Discrete Diffusion in Large Language and Multimodal Models: A Surveycited by 61
  11. 2025dParallel: Learnable Parallel Decoding for dLLMscited by 59
  12. 2025LLaDA-MoE: A Sparse MoE Diffusion Language Modelcited by 54
38 more
  1. 2025Any-Order Flexible Length Masked Diffusioncited by 48
  2. 2025Wide-In, Narrow-Out: Revokable Decoding for Efficient and Effective DLLMscited by 45
  3. 2025Diffusion Language Models are Super Data Learnerscited by 45
  4. 2025KLASS: KL-Guided Fast Inference in Masked Diffusion Modelscited by 44
  5. 2025Accelerating Diffusion Large Language Models with SlowFast Sampling: The Three Golden Principlescited by 43
  6. 2025ParallelBench: Understanding the Trade-offs of Parallel Decoding in Diffusion LLMscited by 42
  7. 2025LongLLaDA: Unlocking Long Context Capabilities in Diffusion LLMscited by 38
  8. 2025DPad: Efficient Diffusion Language Models with Suffix Dropoutcited by 38
  9. 2025SPG: Sandwiched Policy Gradient for Masked Diffusion Language Modelscited by 38
  10. 2025KVCOMM: Online Cross-context KV-cache Communication for Efficient LLM-based Multi-agent Systemscited by 35
  11. 2025Plan for Speed: Dilated Scheduling for Masked Diffusion Language Modelscited by 34
  12. 2026d3LLM: Ultra-Fast Diffusion LLM using Pseudo-Trajectory Distillationcited by 34
  13. 2025Esoteric Language Models: A Family of Any-Order Diffusion LLMscited by 33
  14. 2025dInfer: An Efficient Inference Framework for Diffusion Language Modelscited by 33
  15. 2025Beyond Fixed: Training-Free Variable-Length Denoising for Diffusion Large Language Modelscited by 32
  16. 2025Attention Is All You Need for KV Cache in Diffusion LLMscited by 32
  17. 2025AdaBlock-dLLM: Semantic-Aware Diffusion LLM Inference via Adaptive Block Sizecited by 30
  18. 2025Principled RL for Diffusion LLMs Emerges from a Sequence-Level Perspectivecited by 30
  19. 2025Efficient-DLM: From Autoregressive to Diffusion Language Models, and Beyond in Speedcited by 30
  20. 2025Self Speculative Decoding for Diffusion Large Language Modelscited by 29
  21. 2025Learning Unmasking Policies for Diffusion Language Modelscited by 29
  22. 2025d2: Improving Reasoning in Diffusion Language Models via Trajectory Likelihood Estimationcited by 27
  23. 2026ELF: Embedded Language Flowscited by 27
  24. 2025Learning to Parallel: Accelerating Diffusion Large Language Models via Learnable Parallel Decodingcited by 26
  25. 2025Jailbreaking Large Language Diffusion Models: Revealing Hidden Safety Flaws in Diffusion-Based Text Generationcited by 24
  26. 2025d2Cache: Accelerating Diffusion-Based LLMs via Dual Adaptive Cachingcited by 24
  27. 2025Set Block Decoding is a Language Model Inference Acceleratorcited by 22
  28. 2025Diffusion LLM with Native Variable Generation Lengths: Let [EOS] Lead the Waycited by 22
  29. 2025From Bits to Rounds: Parallel Decoding with Exploration for Diffusion Language Modelscited by 22
  30. 2025From Next-Token to Next-Block: A Principled Adaptation Path for Diffusion LLMscited by 22
  31. 2025Quantization Meets dLLMs: A Systematic Study of Post-training Quantization for Diffusion LLMscited by 21
  32. 2025Structuring The Future: Diffusion LLM Speculative Decoding via Calibrated Draft Graphscited by 21
  33. 2025Training Optimal Large Diffusion Language Modelscited by 21
  34. 2025Accelerating Diffusion LLM Inference via Local Determinism Propagationcited by 21
  35. 2025Encoder-Decoder Diffusion Language Models for Efficient Training and Inferencecited by 21
  36. 2025CDLM: Consistency Diffusion Language Models For Faster Samplingcited by 20
  37. 2026LocateAnything: Fast and High-Quality Vision-Language Grounding with Parallel Box Decodingcited by 20
  38. 2025Empirical Analysis of Decoding Biases in Masked Diffusion Modelscited by 19

50 most cited of 404 All 404 (JSON)

Similar, not cited

Close in wording, no citation either way

  1. 2025Accelerating Diffusion LLMs via Adaptive Parallel Decoding
  2. 2026Unlocking Lossless Speedups in LLMs via Discrete Diffusion
  3. 2025FlashDLM: Accelerating Diffusion Language Model Inference via Efficient KV Caching and Guided Diffusion
  4. 2025dKV-Cache: The Cache for Diffusion Language Models
  5. 2025WeDLM: Reconciling Diffusion Language Models with Standard Causal Attention for Fast Inference
  6. 2026Divide and Conquer: Accelerating Diffusion-Based Large Language Models via Adaptive Parallel Decoding
  7. 2025Seed Diffusion: A Large-Scale Diffusion Language Model with High-Speed Inference
  8. 2026Mean-Field Parallel Decoding for Discrete Diffusion Language Models