Speculative Correction: Draft-then-Refine Decoding for Diffusion Language Models

• URL: http://arxiv.org/abs/2608.02625v1
• Date: Tue, 21 Jul 2026 17:58:10 GMT
• Title: Speculative Correction: Draft-then-Refine Decoding for Diffusion Language Models
• Abstract summary: We study a simple plug-and-play inference pattern: first generate a complete draft, then refine the full response using bidirectional diffusion.<n>We evaluate two configurations using LLaDA2.1-Flash and LLaDA2.1-Mini.<n>Flash-Flash improves GSM8K-384 accuracy from 0.848 to 0.899 while running 1.20 times faster than the selected Flash block-autoregressive baseline.<n>Mini-Flash provides useful quality-latency trade-offs, including MATH-384 performance of 0.294 versus 0.300 for Flash while running 2.17 times faster.
• Score: 50.758327861505784
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Diffusion language models (DLMs) can revise tokens bidirectionally, but standard decoding procedures often adapt them to left-to-right generation by producing text block by block. We study a simple plug-and-play inference pattern: first generate a complete draft, then refine the full response using bidirectional diffusion. Using LLaDA2.1-Flash and LLaDA2.1-Mini, we evaluate two configurations. In Flash-Flash, the same Flash model serves as both drafter and refiner, testing whether an existing model can improve its own block-autoregressive output through global refinement. In Mini-Flash, inspired by speculative decoding, we introduce speculative correction: Mini drafts a full response, and Flash revises it as an editable initialization. Flash-Flash improves GSM8K-384 accuracy from 0.848 to 0.899 while running 1.20 times faster than the selected Flash block-autoregressive baseline, and improves MBPP-384 from 0.545 to 0.693. Latency-window-matched Flash-only controls indicate that these gains persist after targeted tuning of block-autoregressive decoding. Causal ablations indicate that completed drafts provide useful initializations: refinement from a fully masked span performs poorly, full global refinement provides a clear additional gain on GSM8K, and local refinement captures much of the gain on MBPP and MATH. Mini-Flash provides useful quality-latency trade-offs, including MATH-384 performance of 0.294 versus 0.300 for Flash while running 2.17 times faster. These results support a Pareto-frontier interpretation rather than the claim that the heterogeneous cascade uniformly matches Flash quality. Overall, same-model draft-and-refine provides evidence that bidirectional refinement is a useful decoding primitive for DLMs, while speculative correction demonstrates a training-free route to fast DLM generation.

Related papers

• DiffusionGemma Technical Report [72.10570774996219]
  We introduce DiffusionGemma, a language model that uses discrete diffusion to generate text at exceptionally high speed.<n>Rather than decoding one token at a time, DiffusionGemma iteratively refines blocks of 256 tokens in parallel.<n>Our compute-efficient two-stage training pipeline uses fewer than 10% of the starting AR model's total training token budget.
  arXiv  Detail & Related papers  (2026-07-31T16:11:46Z)
• WhiFlash: Accelerating Speculative Decoding with Token-Level Cross-Paradigm Routing [44.656073829954636]
  WhiFlash is a cross-paradigm SD method that unifies autoregressive and diffusion-based parallel drafting under a single token-level controller.<n>We show that WhiFlash achieves significantly higher acceptance lengths, yielding category-specific throughput gains of up to 69.6%.
  arXiv  Detail & Related papers  (2026-06-05T13:23:33Z)
• Self-Adversarial One Step Generation via Condition Shifting [11.426065945265647]
  APEX is a framework for adversarial correction signals extracted endogenously from a flow model.<n>Our 0.6B model surpasses FLUX-Schnell 12B (20$times$ more parameters) in one step quality.<n>With LoRA tuning on Qwen-Image 20B, APEX reaches a GenEval score of 0.89 at NFE=1 in 6 hours, surpassing the original 50-step teacher (0.87) and providing a 15.33$times$ inference speedup.
  arXiv  Detail & Related papers  (2026-04-14T05:54:33Z)
• JoyAI-LLM Flash: Advancing Mid-Scale LLMs with Token Efficiency [56.46698214215968]
  JoyAI-LLM Flash is an efficient Mixture-of-Experts (MoE) language model designed to redefine the trade-off between strong performance and token efficiency.<n>JoyAI-LLM Flash is pretrained on a massive corpus of 20 trillion tokens and further optimized through a rigorous post-training pipeline.
  arXiv  Detail & Related papers  (2026-04-03T13:52:38Z)
• FlashSampling: Fast and Memory-Efficient Exact Sampling [62.5203057469482]
  FlashSampling is an exact sampling primitive that fuses sampling into the LM-head matmul and never materializes the logits tensor in.<n>Across H100, H200, B200, and B300 GPUs, FlashSampling speeds up kernel-level decode workloads.<n>In end-to-end vLLM experiments, it reduces time per output token by up to $19%$ on the models we test.
  arXiv  Detail & Related papers  (2026-03-16T19:37:08Z)
• DFlash: Block Diffusion for Flash Speculative Decoding [11.98141750480807]
  Autoregressive large language models (LLMs) deliver strong performance but require inherently sequential decoding.<n>We introduce DFlash, a speculative decoding framework that employs a lightweight block diffusion model for parallel drafting.
  arXiv  Detail & Related papers  (2026-02-05T18:59:30Z)
• Balancing Understanding and Generation in Discrete Diffusion Models [58.62235340638143]
  Masked Diffusion Language Models (MDLM) excel at semantic understanding and zero-shot generalization.<n>Uniform-noise Diffusion Language Models (UDLM) achieve strong few-step generation quality.<n>We propose XDLM, which bridges the two paradigms via a stationary noise kernel.
  arXiv  Detail & Related papers  (2026-02-01T18:00:35Z)
• Fast-dLLM v2: Efficient Block-Diffusion LLM [64.38006546510337]
  Fast-dLLM v2 is a block diffusion language model that adapts pretrained AR models into dLLMs for parallel text generation.<n>This represents a 500x reduction in training data compared to full-attention diffusion LLMs such as Dream (580B tokens)
  arXiv  Detail & Related papers  (2025-09-30T14:40:18Z)
• FlashMLA-ETAP: Efficient Transpose Attention Pipeline for Accelerating MLA Inference on NVIDIA H20 GPUs [2.9406946721643092]
  This paper introduces FlashMLA-ETAP, a novel framework that enhances MLA inference for the single-instance deployment scenario.<n>ETAP reconfigures attention computation through transposition to align the KV context length with the (M)-dimension in WGMMA operations.<n>FlashMLA-ETAP achieves a 2.78x speedup over FlashMLA at 64K sequence length (batch size 16), with 5.24x and 4.94x improvements over FlashAttention-3 and FlashInfer, respectively.
  arXiv  Detail & Related papers  (2025-05-13T17:45:34Z)
• FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning [11.508362885430133]
  We exploit the asymmetric GPU memory hierarchy to bring significant memory saving and runtime speedup. FlashAttention is still not nearly as fast as optimized matrix-multiply (GEMM) operations, reaching only 25-40% of the theoretical maximum FLOPs/s. We propose FlashAttention-2, with better work partitioning to address these issues.
  arXiv  Detail & Related papers  (2023-07-17T17:50:36Z)

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.