Diffusion-LLM-Papers

We’ve collected a series of papers on Diffusion Language Models.

Since our time is limited and we can't cover everything, please feel free to submit a pull request to contribute.

Theoretical Basis

Deep Unsupervised Learning using Nonequilibrium Thermodynamics
2015-3-12, Paper

Structured Denoising Diffusion Models in Discrete State-Spaces
2021-7-7, Paper

Discrete Diffusion Modeling by Estimating the Ratios of the Data Distribution
2023-10-25, Paper

Your Absorbing Discrete Diffusion Secretly Models the Conditional Distributions of Clean Data
2024-6-6, Paper

Simplified and Generalized Masked Diffusion for Discrete Data
2024-6-6, Paper

Simple and Effective Masked Diffusion Language Models
2024-6-11, Paper

Foundation Model

LLaDA: Large Language Diffusion Models
2025-2-14, Paper

Dream 7B
2025-4-2, Paper

Seed Diffusion: A Large-Scale Diffusion Language Model with High-Speed Inference
2025-8-4, Paper

Multimodal Model

Multimodal Understanding

LLaDA-V: Large Language Diffusion Models with Visual Instruction Tuning
2025-5-22, Paper

LaViDa: A Large Diffusion Language Model for Multimodal Understanding
2025-5-22, Paper

Dimple: Discrete Diffusion Multimodal Large Language Model with Parallel Decoding
2025-5-22, Paper

Unified Multimodal Model

MMaDA: Multimodal Large Diffusion Language Models
2025-5-21, Paper

Speech / ASR

Whisfusion: Parallel ASR Decoding via a Diffusion Transformer
2025-8-9, Paper

Fast Sampling

Enable KV Cache

dKV-Cache: The Cache for Diffusion Language Models
2025-5-21, Paper

dLLM-Cache: Accelerating Diffusion Large Language Models with Adaptive Caching
2025-5-22, Paper

Accelerating Diffusion Language Model Inference via Efficient KV Caching and Guided Diffusion
2025-5-27, Paper

Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding
2025-5-28, Paper

Esoteric Language Models
2025-6-2, Paper

Sparse-dLLM: Accelerating Diffusion LLMs with Dynamic Cache Eviction
2025-8-4, Paper

DPad: Efficient Diffusion Language Models with Suffix Dropout
2025-8-19, Paper

Advanced Sampling Method

Dimple: Discrete Diffusion Multimodal Large Language Model with Parallel Decoding
2025-5-22, Paper

Variational Autoencoding Discrete Diffusion with Enhanced Dimensional Correlations Modeling
2025-5-23, Paper

Accelerating Diffusion Language Model Inference via Efficient KV Caching and Guided Diffusion
2025-5-27, Paper

Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding
2025-5-28, Paper

Accelerated Sampling from Masked Diffusion Models via Entropy Bounded Unmasking
2025-5-30, Paper

Accelerating Diffusion LLMs via Adaptive Parallel Decoding
2025-5-31, Paper

Accelerating Diffusion Large Language Models with SlowFast Sampling: The Three Golden Principles
2025-6-12, Paper

Wide-In, Narrow-Out: Revokable Decoding for Efficient and Effective DLLMs
2025-7-24，Paper

DPad: Efficient Diffusion Language Models with Suffix Dropout
2025-8-19, Paper

Reinforcement Learning

d1: Scaling Reasoning in Diffusion Large Language Models via Reinforcement Learning
2025-4-16, Paper

Reinforcing the Diffusion Chain of Lateral Thought with Diffusion Language Models
2025-5-15, Paper

LLaDA 1.5: Variance-Reduced Preference Optimization for Large Language Diffusion Models
2025-5-25, Paper

DiffuCoder: Understanding and Improving Masked Diffusion Models for Code Generation
2025-7-25, Paper

MDPO: Overcoming the Training-Inference Divide of Masked Diffusion Language Models
2025-8-18, Paper

Revolutionizing Reinforcement Learning Framework for Diffusion Large Language Models
2025-9-8, Paper

Inpainting-Guided Policy Optimization for Diffusion Large Language Models
2025-9-12, Paper

Long Context

LongLLaDA: Unlocking Long Context Capabilities in Diffusion LLMs
2025-6-17, Paper

Variable Length

Edit Flows: Flow Matching with Edit Operations
2025-6-10, Paper

DreamOn: Diffusion Language Models For Code Infilling Beyond Fixed-Size Canvas
2025-7-15, Paper

Beyond Fixed: Variable-Length Denoising for Diffusion Large Language Models
2025-8-4, Paper

Any-Order Flexible Length Masked Diffusion
2025-8-31, Paper

Others

Time Is a Feature: Exploiting Temporal Dynamics in Diffusion Language Models
2025-8-12, Paper

Thinking Inside the Mask: In-Place Prompting in Diffusion LLMs
2025-8-14, Paper