model · first seen 4 Oct 2026
diffusion language model
Also: DLM, DLLM, MDLM, diffusion LLM, masked diffusion, diffusion adapters, discrete diffusion, masked diffusion LM
Language model that generates text by iteratively denoising masked or noisy token sequences.
Diffusion language models generate text by starting from a masked or noisy token sequence and refining it over several denoising steps, rather than producing one token at a time from left to right. Most of the sources are arXiv papers that build on open models such as LLaDA and Dream. They address faster decoding through speculative branches and attention pruning, better-calibrated choice of which tokens to commit, remasking and self-correction, and post-training with ELBO surrogates. Practical activity is appearing too: a Mojo repo for local CPU inference, a 35B mixture-of-experts diffusion model aimed at agentic coding, and a Hacker News thread on Inception Labs' Mercury 3.
If the remaining problems are solved (inference cost per step, miscalibrated confidence, and decoding that drifts back to pseudo-autoregressive order), parallel generation could give a real latency and throughput alternative to autoregressive LLMs, especially for code. The spread from papers into released weights and local inference tooling suggests the approach is moving beyond pure research.
Analyst note (emerging), 17 Nov 2026.
Mentions per week
PapersCodeDiscussion
Rank 7 on 16 Nov 2026 · 72 mentions in 28 days · 69 independent groups · papers → code → discussion · first seen 6 weeks ago · acceleration -0.03
Earliest sources
- SpecFold: Folding Multi-Branch Redundancy for Faster Speculative Decoding in Diffusion Language ModelsPapers · 4 Oct 2026 · Chung-En Ho, Weiyu Sun, Cheng-Jhih Shih, He Li, Yong Liu, Yingyan Celine Lin
- Bayesian Entropy-based Reordering for Calibrated Diffusion Language ModelsPapers · 4 Oct 2026 · Zhejun Jiang, Mijung Park
- vladimirrott/mojo-dllmCode · 4 Oct 2026 · vladimirrott · 541 stars
- Generate What You Can Trust: Content Credibility in Generative RecommendersPapers · 5 Oct 2026 · Zhuo Cai, Guanghao Wu, Shoujin Wang, Peilin Zhou, Victor W. Chu
- Noise Out, Bias In: Targeted Bias Injection in Diffusion Language Models via Closed-Loop Activation SteeringPapers · 5 Oct 2026 · Sarim Hashmi, Mukul Ranjan, Abdelrahman Elsayed, Muhammad Umer Sheikh, Fahad Sha…
- SOL: Measuring Gaps between Text Distributions by Double Sliced Wasserstein MetricsPapers · 5 Oct 2026 · Gregor Kornhardt, Moritz Piening, Jannis Chemseddine, Gabriele Steidl
Latest sources
- K2-Horizon: A Frontier Fleet of Open Source Language ModelsPapers · 16 Nov 2026 · K2 Team, Zhengzhong Liu, Haonan Li, Xuezhe Ma, Alexander Moreno, Subham Sekhar S…
- Fast at the Frontier, Wrong at the Fallback: A Commit-Level Analysis of Threshold Decoding in Diffusion Language ModelsPapers · 16 Nov 2026 · Roman Makarov, Zhe Xu, Jian Kang
- PlanDiv: Test-Time Scaling with Latent Plans for Vision-Language-Action ModelsPapers · 15 Nov 2026 · Hanbo Zhang, Lianlei Shan, Liuhaichen Yang, Zezhi Tang
- SemMask: Semantic-Prioritized Progressive Masking for Diffusion Language ModelsPapers · 14 Nov 2026 · Qi Chen, Zifeng Cheng, Zhiwei Jiang, Qing Gu
- Empowering Masked Diffusion Models to Self-Correct with Leave-One-Out TransformersPapers · 14 Nov 2026 · Sofian Zalouk, Vincent Counathe, Paul Jünger, Daniel Cao, Daniel Zlotnick, Kili…
- See Every Block Once: Variance-Reduced Post-Training for Masked Diffusion Language ModelsPapers · 13 Nov 2026 · Yunzhen He, Hiroaki Yamagiwa, Hidetoshi Shimodaira
- Learning When Diffusion Language Models Should RegeneratePapers · 12 Nov 2026 · Minhae Oh, Jeonghye Kim, Jinwoo Koo, Yeongoon Kim, Jimin Seo, Minsik Seol, Jungw…
- DRAMA: Dissecting Attention Redundancy for Accelerating Multimodal Diffusion Large Language ModelsPapers · 12 Nov 2026 · Yi-Hsin Hung, Fangfu Liu, Shao-Yuan Lo, Chu-Song Chen
- Schedule-Calibrated Commitment Order in Blockwise Masked Diffusion Language ModelsPapers · 11 Nov 2026 · Gaurav Sarkar, Umang Pandey, Pratik Prabhanjan Brahma, Zicheng Liu, Emad Barsoum
- Mask-Pattern Position Augmentation: Training-Free Long-Context Extrapolation for Diffusion LLMsPapers · 11 Nov 2026 · Gaurav Sarkar, Syed Affan Daimi
- Blurring the Block Boundaries: Overlapping Block Denoising for Efficient Diffusion Language ModelsPapers · 10 Nov 2026 · Amir Mann, Jonathan Fhima, Shaked Brody, Or Avitan, Asif Zamir, Shahar Tsiper, N…
- When Should Diffusion Language Models Reconsider Their Commitments?Papers · 8 Nov 2026 · Haoyang Zheng, Yidong Ouyang, Blake Jianhang Chen, Tian Tan, Ziqian Lin, Liqian…