A curated collection of papers, models, and resources for top-tier researches of diffusion large language models.
Note
This repository is proudly maintained by the frontline research mentors at QuenithAI (应达学术). It aims to provide the most comprehensive and cutting-edge map of papers and technologies in the field of Diffusion Large Language Models.
Your contributions are also vital—feel free to open an issue or submit a pull request to become a collaborator of this repository. We expect your participation!
If you require expert 1-on-1 guidance on your submissions to top-tier conferences and journals, we invite you to contact us via WeChat or E-mail.
本仓库由 「应达学术」(QuenithAI) 的一线科研导师团队倾力打造并持续维护，旨在为您呈现文生图领域最全面、最前沿的论文。
您的贡献对我们和社区来说至关重要——我们诚邀有志之士通过 open an issue 或 submit a pull request 来成为这个项目的合作者之一，期待您的加入！
⚡ Latest Updates
- (Sep 17th, 2025): Initial update of the repository.
- 📚 Table of Contents
- ✍️ Survey Papers
- 📜 Papers & Models
- 🎓 About Us
- 🤝 Contributing
- 💬 Join the Community
- Discrete Diffusion in Large Language and Multimodal Models: A Survey
- A Survey on Diffusion Language Models
-
[CVPR 2026] LLaDA-V: Large Language Diffusion Models with Visual Instruction Tuning
-
[CVPR 2026] LLaDA-MedV: Exploring Large Language Diffusion Models for Biomedical Image Understanding
-
[CVPR 2026] Sparse-LaViDa: Sparse Multimodal Discrete Diffusion Language Models
-
[CVPR 2026] dMLLM-TTS: Self-Verified and Efficient Test-Time Scaling for Diffusion Multi-Modal Large Language Models
-
[ICLR 2026] Any-Order Flexible Length Masked Diffusion
-
[ICLR 2026] Attention Is All You Need for KV Cache in Diffusion LLMs
-
[ICLR 2026] Beyond Fixed: Training-Free Variable-Length Denoising for Diffusion Large Language Models
-
[ICLR 2026] DPad: Efficient Diffusion Language Models with Suffix Dropout
-
[ICLR 2026] DiffuCoder: Understanding and Improving Masked Diffusion Models for Code Generation
-
[ICLR 2026] Diffusion LLMs Can Do Faster-Than-AR Inference via Discrete Diffusion Forcing
-
[ICLR 2026] Diffusion Language Model Knows the Answer Before It Decodes
-
[ICLR 2026] Diffusion Language Models For Code Infilling Beyond Fixed-size Canvas
-
[ICLR 2026] Fast-dLLM v2: Efficient Block-Diffusion LLM
-
[ICLR 2026] Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding
-
[ICLR 2026] FlashDLM: Accelerating Diffusion Language Model Inference via Efficient KV Caching and Guided Diffusion
-
[ICLR 2026] Improving Reasoning for Diffusion Language Models via Group Diffusion Policy Optimization
-
[ICLR 2026] Inpainting-Guided Policy Optimization for Diffusion Large Language Models
-
[ICLR 2026] Learning to Parallel: Accelerating Diffusion Large Language Models via Adaptive Parallel Decoding
-
[ICLR 2026] On the Reasoning Abilities of Masked Diffusion Language Models
-
[ICLR 2026] Quant-dLLM: Post-Training Extreme Low-Bit Quantization for Diffusion Large Language Models
-
[ICLR 2026] Revolutionizing Reinforcement Learning Framework for Diffusion Large Language Models
-
[ICLR 2026] SPG: Sandwiched Policy Gradient for Masked Diffusion Language Models
-
[ICLR 2026] Soft-Masked Diffusion Language Models
-
[ICLR 2026] SparseD: Sparse Attention for Diffusion Language Models
-
[ICLR 2026] Time Is a Feature: Exploiting Temporal Dynamics in Diffusion Language Models
-
[ICLR 2026] UltraLLaDA: Scaling the Context Length to 128K for Diffusion Large Language Models
-
[ICLR 2026] Unveiling the Potential of Diffusion Large Language Model in Controllable Generation
-
[ICLR 2026] dParallel: Learnable Parallel Decoding for dLLMs
-
[ICLR 2026] wd1: Weighted Policy Optimization for Reasoning in Diffusion Language Models
-
[ICLR 2026] A2D: Any-Order, Any-Step Safety Alignment for Diffusion Language Models
-
[ICLR 2026] Accelerating Diffusion Large Language Models with SlowFast Sampling: The Three Golden Principles
-
[ICLR 2026] Beyond Masks: Efficient, Flexible Diffusion Language Models via Deletion-Insertion Processes
-
[ICLR 2026] Beyond Scattered Acceptance: Fast and Coherent Inference for DLMs via Longest Stable Prefixes
-
[ICLR 2026] Diffusion Language Models are Provably Optimal Parallel Samplers
-
[ICLR 2026] Don't Settle Too Early: Self-Reflective Remasking for Diffusion Language Models
-
[ICLR 2026] Dynamic-dLLM: Dynamic Cache-Budget and Adaptive Parallel Decoding for Training-Free Acceleration of Diffusion LLM
-
[ICLR 2026] ES-dLLM: Efficient Inference for Diffusion Large Language Models by Early-Skipping
-
[ICLR 2026] FS-DFM: Fast and Accurate Long Text Generation with Few-Step Diffusion Language Models
-
[ICLR 2026] Hierarchy Decoding: A Training-free Parallel Decoding Strategy for Diffusion Large Language Models
-
[ICLR 2026] Membership Inference Attacks Against Fine-tuned Diffusion Language Models
-
[ICLR 2026] Parallel Multimodal Diffusion Language Models for Thinking-Aware Editing and Generation
-
[ICLR 2026] Planner Aware Path Learning in Diffusion Language Models Training
-
[ICLR 2026] Rainbow Padding: Mitigating Early Termination in Instruction-Tuned Diffusion LLMs
-
[ICLR 2026] ReFusion: A Diffusion Large Language Model with Parallel Autoregressive Decoding
-
[ICLR 2026] Revokable Decoding for Efficient and Effective DLLMs
-
[ICLR 2026] Scaling Behavior of Discrete Diffusion Language Models
-
[ICLR 2026] Semantic-Aware Diffusion LLM Inference With Adaptive Block Size
-
[ICLR 2026] Stopping Computation for Converged Tokens in Masked Diffusion-LM Decoding
-
[ICLR 2026] Test-Time Scaling in Diffusion LLMs via Hidden Semi-Autoregressive Experts
-
[ICLR 2026] TRACEDET: Hallucination Detection from the Decoding Trace of Diffusion Large Language Models
-
[ICLR 2026] Toward Safer Diffusion Language Models: Discovery and Mitigation of Priming Vulnerability
-
[ICLR 2026] Watermarking Diffusion Language Models
-
[ICLR 2026] What Exactly Does Guidance Do in Masked Discrete Diffusion Models
- DyLLM: Efficient Diffusion LLM Inference via Saliency-based Token Selection and Partial Attention
- Skip to the Good Part: Representation Structure & Inference-Time Layer Skipping in Diffusion vs. Autoregressive LLMs
- Stabilizing Reinforcement Learning for Diffusion Language Models
- Evo: Autoregressive-Diffusion Large Language Models with Evolving Balance
- Omni-Diffusion: Unified Multimodal Understanding and Generation with Masked Discrete Diffusion
- Diffusion Language Models Are Natively Length-Aware
- Beyond Scattered Acceptance: Fast and Coherent Inference for DLMs via Longest Stable Prefixes
- Diffusion LLMs can think EoS-by-EoS
- Free Lunch for Pass@k? Low Cost Diverse Sampling for Diffusion Language Models
- Progressive Refinement Regulation for Accelerating Diffusion Language Model Decoding
- Efficient Self-Evaluation for Diffusion Language Models via Sequence Regeneration
- CoDAR: Continuous Diffusion Language Models are More Powerful Than You Think
- Characterizing Memorization in Diffusion Language Models: Generalized Extraction and Sampling Effects
- D3LM: A Discrete DNA Diffusion Language Model for Bidirectional DNA Understanding and Generation
- LFPO: Likelihood-Free Policy Optimization for Masked Diffusion Models
- Exact Likelihood for Masked Diffusion via Deterministic Unmasking Rules
- MetaState: Persistent Working Memory for Discrete Diffusion Language Models
- Reasoning or Rationalization? The Role of Justifications in Masked Diffusion Models for Fact Verification
- Breaking the Factorization Barrier in Diffusion Language Models
- Why Diffusion Language Models Struggle with Truly Parallel (Non-Autoregressive) Decoding?
- Test-Time Scaling with Diffusion Language Models via Reward-Guided Stitching
- Rejection Mixing: Fast Semantic Propagation of Mask Tokens for Efficient DLLM Inference
- dLLM: Simple Diffusion Language Modeling
- TabDLM: Free-Form Tabular Data Generation via Joint Numerical-Language Diffusion
- Self-Purification Mitigates Backdoors in Multimodal Diffusion Language Models
- Adaptation to Intrinsic Dependence in Diffusion Language Models
- Is Your Diffusion Sampler Actually Correct? A Sampler-Centric Evaluation of Discrete Diffusion Language Models
- IDLM: Inverse-distilled Diffusion Language Models
- Prompt Optimization Via Diffusion Language Models
- AnCoder: Anchored Code Generation via Discrete Diffusion Models
- Sink-Aware Pruning for Diffusion Language Models
- One-step Language Modeling via Continuous Denoising
- Discrete Stochastic Localization for Non-autoregressive Generation
- VDLM: Variable Diffusion LMs via Robust Latent-to-Text Rendering
- Scaling Beyond Masked Diffusion Language Models
- MAGE: All-[MASK] Block Already Knows Where to Look in Diffusion LLM
- LaViDa-R1: Advancing Reasoning for Unified Multimodal Diffusion Language Models
- Can I Have Your Order? Monte-Carlo Tree Search for Slot Filling Ordering in Diffusion Language Models
- Continuous Diffusion Models Can Obey Formal Syntax
- T3D: Few-Step Diffusion Language Models via Trajectory Self-Distillation with Direct Discriminative Optimization
- dVoting: Fast Voting for dLLMs
- DICE: Diffusion Large Language Models Excel at Generating CUDA Kernels
- Embedding Inversion via Conditional Masked Diffusion Language Models
- Search or Accelerate: Confidence-Switched Position Beam Search for Diffusion Language Models
- Advancing Block Diffusion Language Models for Test-Time Scaling
- Where-to-Unmask: Ground-Truth-Guided Unmasking Order Learning for Masked Diffusion Language Models
- Efficient and Stable Reinforcement Learning for Diffusion Language Models
- TEAM: Temporal-Spatial Consistency Guided Expert Activation for MoE Diffusion Language Model Acceleration
- TDGNet: Hallucination Detection in Diffusion Language Models via Temporal Dynamic Graphs
- Improving Variable-Length Generation in Diffusion Language Models via Length Regularization
- DLLM Agent: See Farther, Run Faster
- DLLM-Searcher: Adapting Diffusion Large Language Model for Search Agents
- DAWN: Dependency-Aware Fast Inference for Diffusion LLMs
- Diffusion-State Policy Optimization for Masked Diffusion Language Models
- Stopping Computation for Converged Tokens in Masked Diffusion-LM Decoding
- Stop the Flip-Flop: Context-Preserving Verification for Fast Revocable Diffusion Decoding
- DFlash: Block Diffusion for Flash Speculative Decoding
- DSB: Dynamic Sliding Block Scheduling for Diffusion LLMs
- DLM-Scope: Mechanistic Interpretability of Diffusion Language Models via Sparse Autoencoders
- FlashBlock: Attention Caching for Efficient Long-Context Block Diffusion
- EntRGi: Entropy Aware Reward Guidance for Diffusion Language Models
- Swordsman: Entropy-Driven Adaptive Block Partition for Efficient Diffusion Language Models
- UnMaskFork: Test-Time Scaling for Masked Diffusion via Deterministic Action Branching
- Reasoning with Latent Tokens in Diffusion Language Models
- Lookahead Path Likelihood Optimization for Diffusion LLMs
- Step-Wise Refusal Dynamics in Autoregressive and Diffusion Language Models
- SPA-Cache: Singular Proxies for Adaptive Caching in Diffusion Language Models
- AR-MAP: Are Autoregressive Large Language Models Implicit Teachers for Diffusion Large Language Models?
- Focus-dLLM: Accelerating Long-Context Diffusion LLM Inference via Confidence-Guided Context Focusing
- Self-Rewarding Sequential Monte Carlo for Masked Diffusion Language Models
- Prism: Efficient Test-Time Scaling via Hierarchical Search and Self-Verification for Discrete Diffusion Language Models
- Balancing Understanding and Generation in Discrete Diffusion Models
- Dynamic Expert Sharing: Decoupling Memory from Parallelism in Mixture-of-Experts Diffusion LLMs
- Lookahead-then-Verify: Reliable Constrained Decoding for Diffusion LLMs under Context-Free Grammars
- Diffusion LMs Can Approximate Optimal Infilling Lengths Implicitly
- A Fragile Guardrail: Diffusion LLM's Safety Blessing and Its Failure Mode
- FOCUS: DLLMs Know How to Tame Their Compute Bound
- FourierSampler: Unlocking Non-Autoregressive Potential in Diffusion Language Models via Frequency-Guided Generation
- DIFFA-2: A Practical Diffusion Large Language Model for General Audio Understanding
- dgMARK: Decoding-Guided Watermarking for Diffusion Language Models
- Residual Context Diffusion Language Models
- Relaxing Positional Alignment in Masked Diffusion Language Models
- Time-Annealed Perturbation Sampling: Diverse Generation for Diffusion Language Models
- ρ-EOS: Training-free Bidirectional Variable-Length Control for Masked Diffusion LLMs
- Tuning the Implicit Regularizer of Masked Diffusion Language Models: Enhancing Generalization via Insights from k-Parity
- Thinking Out of Order: When Output Order Stops Reflecting Reasoning Order in Diffusion Language Models
- Causal Autoregressive Diffusion Language Model
- ILRR: Inference-Time Steering Method for Masked Diffusion Language Models
- Beyond GEMM-Centric NPUs: Enabling Efficient Diffusion LLM Sampling
- Improving Diffusion Language Model Decoding through Joint Search in Generation Order and Token Space
- Window-Diffusion: Accelerating Diffusion Language Model Inference with Windowed Token Pruning and Caching
- Membership Inference Attacks Against Fine-tuned Diffusion Language Models
- MaDiS: Taming Masked Diffusion Language Models for Sign Language Generation
- Streaming-dLLM: Accelerating Diffusion LLMs via Suffix Pruning and Dynamic Decoding
- dLLM-ASR: A Faster Diffusion LLM-based Framework for Speech Recognition
- VidLaDA: Bidirectional Diffusion Large Language Models for Efficient Video Understanding
- Stable-DiffCoder: Pushing the Frontier of Code Diffusion Large Language Model
- Parallelism and Generation Order in Masked Diffusion Language Models: Limits Today, Potential Tomorrow
- The Flexibility Trap: Why Arbitrary Order Limits Reasoning Potential in Diffusion Language Models
- STEAD: Robust Provably Secure Linguistic Steganography with Diffusion Language Model
- Mechanism Shift During Post-training from Autoregressive to Masked Diffusion Language Models
- Diffusion Large Language Models for Black-Box Optimization
- Top 10 Open Challenges Steering the Future of Diffusion Language Model and Its Variants
- The Bitter Lesson of Diffusion Language Models for Agentic Workflows: A Comprehensive Reality Check
- Plan, Verify and Fill: A Structured Parallel Decoding Approach for Diffusion Language Models
- Unlocking the Potentials of Retrieval-Augmented Generation for Diffusion Language Models: A Semantic Drift Perspective
- T*: Progressive Block Scaling for Masked Diffusion Language Models Through Trajectory Aware Reinforcement Learning
- d3LLM: Ultra-Fast Diffusion LLM using Pseudo-Trajectory Distillation
- Beyond Hard Masks: Progressive Token Evolution for Diffusion Language Models
- DiffER: Diffusion Entity-Relation Modeling for Reversal Curse in Diffusion Large Language Models
- Agents of Diffusion: Enhancing Diffusion Language Models with Multi-Agent Reinforcement Learning for Structured Data Generation (Extended Version)
- BackPlay: Plug-in Look-Back Self-Correction for Diffusion Language Models
- STaRR: Spatial-Temporal Token-Dynamics-Aware Responsive Remasking for Diffusion Language Models
- DIP: Dynamic In-Context Planner For Diffusion Language Models
- CD4LM: Consistency Distillation and aDaptive Decoding for Diffusion Language Models
- Deferred Commitment Decoding for Diffusion Language Models
- DLM-One: Diffusion Language Models for One-Step Sequence Generation
- The Diffusion Duality
- Improving Text Style Transfer using Masked Diffusion Language Models with Inference-time Scaling
- Masked Diffusion Language Models with Frequency-Informed Training
- Beyond Autoregression: An Empirical Study of Diffusion Large Language Models for Code Generation
- Fast and Fluent Diffusion Language Models via Convolutional Decoding and Rejective Fine-tuning
- Spiffy: Multiplying Diffusion LLM Acceleration via Lossless Speculative Decoding
- Taming Masked Diffusion Language Models via Consistency Trajectory Reinforcement Learning with Fewer Decoding Step
- RFG: Test-Time Scaling for Diffusion Large Language Model Reasoning with Reward-Free Guidance
- Training Optimal Large Diffusion Language Models
- What Makes Diffusion Language Models Super Data Learners?
- Self Speculative Decoding for Diffusion Large Language Models
- Understanding the Trade-offs of Parallel Decoding in Diffusion LLMs
- SDAR: A Synergistic Diffusion-AutoRegression Paradigm for Scalable Sequence Generation
- Lumina-DiMOO: An Omni Diffusion Large Language Model for Multi-Modal Generation and Understanding
- Improving Reasoning for Diffusion Language Models via Group Diffusion Policy Optimization
- SPG: Sandwiched Policy Gradient for Masked Diffusion Language Models
- UltraLLaDA: Scaling the Context Length to 128K for Diffusion Large Language Models
- On the Reasoning Abilities of Masked Diffusion Language Models
- Soft-Masked Diffusion Language Models
- MRO: Enhancing Reasoning in Diffusion Language Models via Multi-Reward Optimization
- Encoder-Decoder Diffusion Language Models for Efficient Sequence Generation
- Diffusion LLM with Native Variable Generation Lengths
- Diffuse Thinking: Exploring Diffusion Language Models as Efficient Thought Proposers for Reasoning
- Diffusion LLMs are Natural Adversaries for any LLM
- Reasoning in Diffusion Large Language Models is Concentrated in Dynamic Confusion Zones
- Diffusion Language Models are Super Data Learners
- Attention Sinks in Diffusion Language Models
- Seed Diffusion: A Large-Scale Diffusion Language Model with High-Speed Inference
- Diffusion LLMs Can Do Faster-Than-AR Inference via Discrete Diffusion Forcing
- Attention Is All You Need for KV Cache in Diffusion LLMs
- CoDA: Coding LM via Diffusion Adaptation
- LaViDa: A Large Diffusion Language Model for Multimodal Understanding
- Dimple: Discrete Diffusion Multimodal Large Language Model with Parallel Decoding
- Plan for Speed: Dilated Scheduling for Masked Diffusion Language Models
- Sequential Diffusion Language Models
- LLaDA-MoE: A Sparse MoE Diffusion Language Model
- Sparse-dLLM: Accelerating Diffusion LLMs with Dynamic Cache Eviction
- Diffusion Language Models Know the Answer Before Decoding
- Learning to Parallel: Accelerating Diffusion Large Language Models via Adaptive Parallel Decoding
- d^2Cache: Accelerating Diffusion-Based LLMs via Dual Adaptive Caching
- DPad: Efficient Diffusion Language Models with Suffix Dropout
- dKV-Cache: The Cache for Diffusion Language Models
- Accelerating Diffusion Language Model Inference via Efficient KV Caching and Guided Diffusion
- Time Is a Feature: Exploiting Temporal Dynamics in Diffusion Language Models
- Set Block Decoding is a Language Model Inference Accelerator
- Quantization Meets dLLMs: A Systematic Study of Post-training Quantization for Diffusion LLMs
- CtrlDiff: Boosting Large Diffusion Language Models with Dynamic Block Prediction and Controllable Generation
- Dream 7B: Diffusion Large Language Models
- d1: Scaling Reasoning in Diffusion Large Language Models via Reinforcement Learning
- Reinforcing the Diffusion Chain of Lateral Thought with Diffusion Language Models
- LLaDA 1.5: Variance-Reduced Preference Optimization for Large Language Diffusion Models
- MDPO: Overcoming the Training-Inference Divide of Masked Diffusion Language Models
- DIFFA: Large Language Diffusion Models Can Listen and Understand
- The Devil behind the mask: An emergent safety vulnerability of Diffusion LLMs
- Whisfusion: Parallel ASR Decoding via a Diffusion Transformer
- LLaDA-VLA: Vision Language Diffusion Action Models
- Large Language Diffusion Models
QuenithAI is a professional organization composed of top researchers, dedicated to providing high-quality 1-on-1 research mentoring for university students worldwide. Our mission is to help students bridge the gap from theoretical knowledge to cutting-edge research and publish their work in top-tier conferences and journals.
Maintaining this Awesome Text-to-Image Generation list requires significant effort, just as completing a high-quality paper requires focused dedication and expert guidance. If you're looking for one-on-one support from top scholars on your own research project, to quickly identify innovative ideas and make publications, we invite you to contact us ASAP.
➡️ Contact us via WeChat or E-mail to start your research journey.
「应达学术」(QuenithAI) 是一家由顶尖研究者组成，致力于为全球高校学生提供高质量1V1科研辅导的专业机构。我们的使命是帮助学生培养出色卓越的科研技能，在顶级会议和期刊上发表自己的成果。
维护一个GitHub调研仓库需要巨大的精力，正如完成一篇高质量的论文一样，离不开专注的投入和专业的指导。如果您希望在自己的研究项目中，获得来自顶尖学者的一对一支持，我们诚邀您与我们取得联系。
➡️ 欢迎通过 微信 或 邮件 联系我们，开启您的科研之旅。
Contributions are welcome! Please see our Contribution Guidelines for details on how to add new papers, correct information, or improve the repository.
Join our community to stay up-to-date with the latest advancements, share your work, and collaborate with other researchers and developers in the field of video generation, diffusion large language models, and more!
If you are interested, please contact our administrator to join the group.