S2D2: Fast Decoding for Diffusion LLMs via Training-Free Self-Speculation

• URL: http://arxiv.org/abs/2603.25702v1
• Date: Thu, 26 Mar 2026 17:48:50 GMT
• Title: S2D2: Fast Decoding for Diffusion LLMs via Training-Free Self-Speculation
• Abstract summary: Block-diffusion language models combine block-wise autoregressive decoding with within-block parallel denoising.<n>Existing approaches that address this issue either require additional training or incur extra test-time compute.<n>We present S2D2, a training-free self-speculative decoding framework for block-diffusion language models.
• Score: 22.303253139413286
• License: http://creativecommons.org/licenses/by-nc-nd/4.0/
• Abstract: Block-diffusion language models offer a promising path toward faster-than-autoregressive generation by combining block-wise autoregressive decoding with within-block parallel denoising. However, in the few-step regime needed for practical acceleration, standard confidence-thresholded decoding is often brittle: aggressive thresholds hurt quality, while conservative thresholds require unnecessary denoising steps. Existing approaches that address this issue either require additional training or incur extra test-time compute. We present S2D2, a training-free self-speculative decoding framework for block-diffusion language models. Our key observation is that a block-diffusion model becomes autoregressive when the block size is reduced to one, allowing the same pretrained model to act as both drafter and verifier. S2D2 inserts a speculative verification step into standard block-diffusion decoding and uses lightweight routing policies to decide when verification is worth its cost. This yields a hybrid decoding trajectory in which diffusion proposes tokens in parallel, while the autoregressive mode acts as a local sequence-level critic. Across three mainstream block-diffusion families, S2D2 consistently improves the accuracy-speed tradeoff over strong confidence-thresholding baselines. On SDAR, we observe up to $4.7\times$ speedup over autoregressive decoding, and up to $1.57\times$ over a tuned dynamic decoding baseline while improving accuracy by up to $4.5$ points. On LLaDA2.1-Mini, S2D2 remains complementary to built-in self-correction, including a conservative setting where it is $4.4\times$ faster than the static baseline with slightly higher accuracy.

Related papers

• Early-Bird Decoding: Accelerating Diffusion LLMs with Learnable Block Sizes and Parallel Sampling [11.823547220829083]
  Diffusion large language models (dLLMs) offer a promising parallel decoding paradigm as an alternative to autoregressive generation.<n>We propose an "early-bird (EB)" decoding framework, motivated by the observation that tokens with similarly low entropy tend to cluster.<n>EB-Decode 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; and (2) a position-aware sampler that learns to unmask tokens in parallel using fewer decoding steps within predicted variable-length blocks.
  arXiv  Detail & Related papers  (2026-09-15T00:12:37Z)
• HEAT: Faster Fully Homomorphic Inference via Approximations-Weights Co-Adaptation [56.14416807766864]
  Homomorphic encryption allows a server to run a language model directly on encrypted user prompts.<n>Existing approaches fix the iteration counts uniformly across the model rather than tailoring them to each site's error tolerance.<n>We introduce Homomorphic Encryption-Aware Training (HEAT), a fine-tuning method that makes the per-nonline iteration counts learnable.
  arXiv  Detail & Related papers  (2026-09-01T18:02:02Z)
• Trajectory-Level Speculative Decoding for Diffusion Language Models [16.97093186372498]
  Diffusion-based language models (dLLMs) enable parallel token generation through iterative denoising, but existing decoding strategies collapse to single-token generation under low confidence.<n>We develop a trajectory-level speculative framework that constructs draft denoising trajectories via confidence-stratified tree exploration.<n>Our framework reduces denoising by 30-40% and increases tokens-per-step from 2.6 to 4.3, achieving 7-14x speedup over vanilla dLLMs and 1.3x over Fast-dLLM with less than 1% accuracy change across reasoning and code benchmarks.
  arXiv  Detail & Related papers  (2026-08-27T09:42:20Z)
• FlowBlock: Wavefront-Parallel Decoding for Self-Correcting Diffusion Language Models [16.404926372068047]
  Block-wise diffusion large language models (dLLMs) decode sequentially at the block level, enabling effective KV-cache reuse across blocks but making inter-block decoding strictly serial.<n>We propose textbfflowblock, a training-free parallel decoding framework built on two mechanisms.<n>It improves tokens per second (TPS) over LLaDA-2.1 and LLaDA-2.0, two serial block-wise dLLMs, by up to 2.95$times$ and 4.01$times$, while reducing latency by up to 53.6% and 77.1%
  arXiv  Detail & Related papers  (2026-07-20T08:05:03Z)
• Prefix-Adaptive Block Diffusion for Efficient Document Recognition [52.15352911463151]
  Block Diffusion Models (BDMs) support parallel generation, flexible-length output, and KV caching, making them promising for efficient document parsing.<n>We propose the Prefix- Block Diffusion Model (PA-BDM) which replaces intra-block bidirectional denoising with causal denoising from prefix to suffix.<n> Experiments show that the 3B PA-BDM achieves higher recognition scores on several benchmarks and improves inference throughput by 71.6% over the 2.5B MinerU-Diffusion.
  arXiv  Detail & Related papers  (2026-05-16T07:50:13Z)
• SpecDiff-2: Scaling Diffusion Drafter Alignment For Faster Speculative Decoding [48.96349422252313]
  Speculative decoding has become the standard approach for accelerating Large Language Model (LLM) inference.<n>It exploits a lossless draft-then-verify procedure to circumvent the latency of autoregressive decoding, achieving impressive speed-ups.<n>This paper proposes SpecDiff-2, a novel framework to jointly address these two bottlenecks.
  arXiv  Detail & Related papers  (2025-11-01T16:12:56Z)
• Saber: An Efficient Sampling with Adaptive Acceleration and Backtracking Enhanced Remasking for Diffusion Language Model [98.35868970993232]
  Diffusion language models (DLMs) are emerging as a powerful and promising alternative to the dominant autoregressive paradigm.<n>We introduce efficient Sampling with Adaptive acceleration and Backtracking Enhanced Remasking (i.e., Saber) to achieve better inference speed and output quality in code generation.
  arXiv  Detail & Related papers  (2025-10-20T23:38:12Z)
• AdaBlock-dLLM: Semantic-Aware Diffusion LLM Inference via Adaptive Block Size [7.442463267121892]
  Diffusion-based large language models (dLLMs) are gaining attention for their inherent capacity for parallel decoding.<n>This paper presents the first systematic investigation challenging the fixed block size assumption in semi-AR decoding.<n>We introduce AdaBlock-dLLM, a training-free, plug-and-play scheduler that adaptively aligns block boundaries with semantic steps by adjusting block size during runtime.
  arXiv  Detail & Related papers  (2025-09-30T15:53:56Z)
• R-Stitch: Dynamic Trajectory Stitching for Efficient Reasoning [80.104336426172]
  Chain-of-thought (CoT) enhances problem-solving ability of large language models.<n>CoT incurs substantial inference cost due to long autoregressive trajectories.<n>We introduce R-Stitch, a training-free hybrid decoding framework.
  arXiv  Detail & Related papers  (2025-07-23T08:14:36Z)
• Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding [51.711605076319216]
  Diffusion-based large language models (Diffusion LLMs) have shown promise for non-autoregressive text generation with parallel decoding capabilities.<n>We introduce a novel block-wise approximate KV Cache mechanism tailored for bidirectional diffusion models, enabling cache reuse with negligible performance drop.<n>We propose a confidence-aware parallel decoding strategy that selectively decodes tokens exceeding a confidence threshold, mitigating dependency violations and maintaining generation quality.
  arXiv  Detail & Related papers  (2025-05-28T17:39:15Z)
• FlashDLM: Accelerating Diffusion Language Model Inference via Efficient KV Caching and Guided Diffusion [22.207275433870937]
  Diffusion language models offer parallel token generation and inherent bidirectionality.<n>State-of-the-art diffusion models (e.g., Dream 7B, LLaDA 8B) suffer from slow inference.<n>We introduce Guided Diffusion, a training-free method that uses a lightweight pretrained autoregressive model to supervise token unmasking.
  arXiv  Detail & Related papers  (2025-05-27T17:39:39Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.