mlsanityData to 14 Oct 2026

Early-Bird Decoding: Accelerating Diffusion LLMs with Learnable Block Sizes and Parallel Sampling

arXiv 2609.16450 ↗First posted 15 Sep 2026cs.CL cs.AI cs.LGcited by 0

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

Abstract (arXiv)

Diffusion large language models (dLLMs) offer a promising parallel decoding paradigm as an alternative to autoregressive generation through iterative unmasking. However, dLLMs typically require many steps before token confidence reaches the decoding threshold, resulting in inefficient inference even with block-wise KV caching. To accelerate dLLM inference, we for the first time propose an "early-bird (EB)" decoding framework, motivated by the observation that tokens with similarly low entropy tend to cluster and can be jointly decoded earlier, before reaching the confidence threshold. In particular, our EB-Decode framework integrates two key enablers: (1) a learnable network that adaptively groups tokens with similar uncertainty into variable-length blocks, rather than relying on fixed block sizes; (2) a position-aware sampler that learns to unmask tokens in parallel using fewer decoding steps within predicted variable-length blocks. Both components are developed without modifying pretrained dLLM weights and can therefore be directly deployed as plug-ins during serving, with negligible training and inference overhead. Extensive experiments across three models and four benchmarks consistently validate our observation and the effectiveness of EB-Decode, achieving 3.53-18.76× higher throughput than the vanilla decoding method and up to 1.58× higher throughput over the strongest baseline, Fast-dLLM, with comparable accuracy.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Diffusion LLM and MoE inference, one of 12 directions in Efficient LLM inference and serving. Builds on 57 of its 62 references that are in the map. First posted on arXiv 15 Sep 2026.

Builds on

57 of 62 references.

Closest ancestors · 10

Same direction, newest first

  1. 2026Why Diffusion Language Models Struggle with Truly Parallel (Non-Autoregressive) Decoding?26 Feb
  2. 2026DSB: Dynamic Sliding Block Scheduling for Diffusion LLMs5 Feb
  3. 2025ParallelBench: Understanding the Trade-offs of Parallel Decoding in Diffusion LLMs6 Oct
  4. 2025dParallel: Learnable Parallel Decoding for dLLMs30 Sep
  5. 2025AdaBlock-dLLM: Semantic-Aware Diffusion LLM Inference via Adaptive Block Size30 Sep
  6. 2025Learning to Parallel: Accelerating Diffusion Large Language Models via Learnable Parallel Decoding29 Sep
  7. 2025Diffusion LLMs Can Do Faster-Than-AR Inference via Discrete Diffusion Forcing8 Aug
  8. 2025Accelerating Diffusion LLMs via Adaptive Parallel Decoding31 May
  9. 2025Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding28 May
  10. 2025dKV-Cache: The Cache for Diffusion Language Models21 May

Other references · 47

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,459
  3. 2021Training Verifiers to Solve Math Word Problemscited by 10,109
  4. 2025Qwen3 Technical Reportcited by 8,820
  5. 2021Evaluating Large Language Models Trained on Codecited by 8,320
  6. 2023Efficient Memory Management for Large Language Model Serving with PagedAttentioncited by 5,944
  7. 2023Let's Verify Step by Stepcited by 3,789
  8. 2022FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awarenesscited by 3,673
  9. 2021Program Synthesis with Large Language Modelscited by 3,637
  10. 2018The Lottery Ticket Hypothesis: Finding Sparse, Trainable Neural Networkscited by 2,262
  11. 2021Structured Denoising Diffusion Models in Discrete State-Spacescited by 1,719
  12. 2025Large Language Diffusion Modelscited by 912
35 more
  1. 2024Simple and Effective Masked Diffusion Language Modelscited by 848
  2. 2023Discrete Diffusion Modeling by Estimating the Ratios of the Data Distributioncited by 745
  3. 2017Program Induction by Rationale Generation : Learning to Solve and Explain Algebraic Word Problemscited by 650
  4. 2024Simplified and Generalized Masked Diffusion for Discrete Datacited by 626
  5. 2017BranchyNet: Fast Inference via Early Exiting from Deep Neural Networkscited by 596
  6. 2025Dream 7B: Diffusion Large Language Modelscited by 544
  7. 2016Adaptive Computation Time for Recurrent Neural Networkscited by 506
  8. 2025Block Diffusion: Interpolating Between Autoregressive and Diffusion Language Modelscited by 409
  9. 2020DeeBERT: Dynamic Early Exiting for Accelerating BERT Inferencecited by 332
  10. 2022Confident Adaptive Language Modelingcited by 310
  11. 2020BERT Loses Patience: Fast and Robust Inference with Early Exitcited by 273
  12. 2025Mercury: Ultra-Fast Language Models Based on Diffusioncited by 235
  13. 2024Masked Diffusion Models are Secretly Time-Agnostic Masked Models and Exploit Inaccurate Categorical Samplingcited by 208
  14. 2025Remasking Discrete Diffusion Models with Inference-Time Scalingcited by 189
  15. 2025Seed Diffusion: A Large-Scale Diffusion Language Model with High-Speed Inferencecited by 178
  16. 2025dLLM-Cache: Accelerating Diffusion Large Language Models with Adaptive Cachingcited by 169
  17. 2019Drawing Early-Bird Tickets: Towards More Efficient Training of Deep Networkscited by 160
  18. 2025SDAR: A Synergistic Diffusion-AutoRegression Paradigm for Scalable Sequence Generationcited by 129
  19. 2020EarlyBERT: Efficient BERT Training via Early-bird Lottery Ticketscited by 79
  20. 2024Beyond Autoregression: Fast LLMs via Self-Distillation Through Timecited by 78
  21. 2025FlashDLM: Accelerating Diffusion Language Model Inference via Efficient KV Caching and Guided Diffusioncited by 77
  22. 2025Diffusion Language Models Know the Answer Before Decodingcited by 70
  23. 2024Distillation of Discrete Diffusion through Dimensional Correlationscited by 48
  24. 2025Wide-In, Narrow-Out: Revokable Decoding for Efficient and Effective DLLMscited by 46
  25. 2025Accelerating Diffusion Large Language Models with SlowFast Sampling: The Three Golden Principlescited by 45
  26. 2025DPad: Efficient Diffusion Language Models with Suffix Dropoutcited by 39
  27. 2025dInfer: An Efficient Inference Framework for Diffusion Language Modelscited by 37
  28. 2025Plan for Speed: Dilated Scheduling for Masked Diffusion Language Modelscited by 34
  29. 2025Time Is a Feature: Exploiting Temporal Dynamics in Diffusion Language Modelscited by 33
  30. 2021Early-Bird GCNs: Graph-Network Co-Optimization Towards More Efficient GCN Training and Inference via Drawing Early-Bird Lottery Ticketscited by 27
  31. 2024Light-PEFT: Lightening Parameter-Efficient Fine-Tuning via Early Pruningcited by 15
  32. 2025How Efficient Are Diffusion Language Models? A Critical Examination of Efficiency Evaluation Practicescited by 10
  33. 2026CD4LM: Consistency Distillation and aDaptive Decoding for Diffusion Language Modelscited by 8
  34. 2025Early-Bird Diffusion: Investigating and Leveraging Timestep-Aware Early-Bird Tickets in Diffusion Models for Efficient Trainingcited by 5
  35. 2026Swordsman: Entropy-Driven Adaptive Block Partition for Efficient Diffusion Language Modelscited by 3

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Fast-dLLM++: Fr'echet Profile Decoding for Faster Diffusion LLM Inference
  2. 2025Fast-dLLM v2: Efficient Block-Diffusion LLM
  3. 2025Free Draft-and-Verification: Toward Lossless Parallel Decoding for Diffusion Large Language Models
  4. 2025From Bits to Rounds: Parallel Decoding with Exploration for Diffusion Language Models
  5. 2026Roll Out and Roll Back: Diffusion LLMs are Their Own Efficiency Teachers
  6. 2025Accelerating Diffusion LLM Inference via Local Determinism Propagation
  7. 2026d3LLM: Ultra-Fast Diffusion LLM using Pseudo-Trajectory Distillation
  8. 2026SimSD: Simple Speculative Decoding in Diffusion Language Models