mlsanityData to 7 Oct 2026

Diffusion Language Model Parallel Decoding via Product-of-Experts Bridge

arXiv 2606.08048 ↗First posted 6 Jun 2026cs.CL cs.LG cs.PFcited by 0

On the map: Diffusion models → Discrete and text diffusion

also in 1 more
Abstract (arXiv)

Diffusion language models (DLMs) offer substantial speed advantages through parallel decoding, but the lack of token dependencies limits generation quality compared to autoregressive (AR) models. Recent progress attempts to bridge the gap via importance sampling, with DLM being the proposal and AR being the target. However, due to the huge gap between their distributions, the sampling requires a large number of particles and is thus expensive to compute. In this paper, we introduce PoE-Bridge, a novel decoding framework that drastically improves generation speed and accuracy by introducing an intermediate distribution to bridge the gap. The distribution is constructed as a Product-of-Experts (PoE) of the DLM proposal and the AR target. With the intermediate distribution, we first use the DLM to draft multiple continuations in parallel, then apply rejection sampling to verify the drafted tokens and move the resulting candidates toward the PoE. We then use importance sampling to further correct the PoE-aligned candidates toward the AR target. We further propose several improved techniques, including mixed-temperature sampling for enhanced diversity and elastic rejection windows for reducing wasted verification. Empirically, PoE-Bridge achieves significantly improved accuracy with 5× speedup over the standard DLM decoding approach, and recovers at least 95% of the target AR model's performance, efficiently advancing most of the quality gap on challenging mathematical reasoning and coding tasks. Our code is available at https://github.com/juntongshi48/poe-bridge.

Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Discrete and text diffusion, one of 14 directions in Diffusion models. Builds on 29 of its 40 references that are in the map. First posted on arXiv 6 Jun 2026.

Builds on

29 of 40 references.

Closest ancestors · 7

Same direction, newest first

  1. 2025Dream 7B: Diffusion Large Language Models21 Aug
  2. 2025DiffuCoder: Understanding and Improving Masked Diffusion Models for Code Generation25 Jun
  3. 2025Large Language Diffusion Models14 Feb
  4. 2024Energy-Based Diffusion Language Models for Text Generation28 Oct
  5. 2024Simple and Effective Masked Diffusion Language Models11 Jun
  6. 2024Simplified and Generalized Masked Diffusion for Discrete Data6 Jun
  7. 2023A Reparameterized Discrete Diffusion Model for Text Generation11 Feb

Other references · 22

  1. 2017Attention Is All You Needcited by 59,932
  2. 2023LLaMA: Open and Efficient Foundation Language Modelscited by 13,958
  3. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
  4. 2021Training Verifiers to Solve Math Word Problemscited by 9,743
  5. 2021Evaluating Large Language Models Trained on Codecited by 8,053
  6. 2021Measuring Mathematical Problem Solving With the MATH Datasetcited by 5,203
  7. 2021Program Synthesis with Large Language Modelscited by 3,509
  8. 2023Code Llama: Open Foundation Models for Codecited by 2,105
  9. 2021Structured Denoising Diffusion Models in Discrete State-Spacescited by 1,665
  10. 2022Fast Inference from Transformers via Speculative Decodingcited by 1,615
  11. 2023Accelerating Large Language Model Decoding with Speculative Samplingcited by 975
  12. 2017Non-Autoregressive Neural Machine Translationcited by 457
10 more
  1. 2022A Continuous Time Framework for Discrete Denoising Modelscited by 437
  2. 2025Accelerating Diffusion LLMs via Adaptive Parallel Decodingcited by 108
  3. 2021Step-unrolled Denoising Autoencoders for Text Generationcited by 102
  4. 2020Fully Non-autoregressive Neural Machine Translation: Tricks of the Tradecited by 99
  5. 2024Speculative Diffusion Decoding: Accelerating Language Generation through Diffusioncited by 57
  6. 2024Faster Cascades via Speculative Decodingcited by 38
  7. 2024ParallelSpec: Parallel Drafter for Efficient Speculative Decodingcited by 20
  8. 2025Accelerating LLM Inference with Lossless Speculative Decoding Algorithms for Heterogeneous Vocabulariescited by 20
  9. 2025Reviving Any-Subset Autoregressive Models with Principled Parallel Sampling and Speculative Decodingcited by 14
  10. 2026Speculative Speculative Decodingcited by 13

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026Rethinking the Generation Order of Block Diffusion Language Models
  2. 2026Locally Coherent Parallel Decoding in Diffusion Language Models
  3. 2026Why Diffusion Language Models Struggle with Truly Parallel (Non-Autoregressive) Decoding?
  4. 2025From Bits to Rounds: Parallel Decoding with Exploration for Diffusion Language Models
  5. 2026Early-Bird Decoding: Accelerating Diffusion LLMs with Learnable Block Sizes and Parallel Sampling
  6. 2025Diffusion Language Models Know the Answer Before Decoding
  7. 2026Introspective Diffusion Language Models
  8. 2026d3LLM: Ultra-Fast Diffusion LLM using Pseudo-Trajectory Distillation