---
title: "On-Policy Self-Distillation for Efficient Diffusion Language Models"
description: "The authors propose Calibrated On-Policy Self-Distillation (COPSD). 34.60% accuracy. 38 references, cited by 1 paper here."
authors: ["Huaisheng Zhu", "MingYu Liu", "Junze Liu", "Zhen Ge", "Tian Wang", "Jiri Gesi", "Dakuo Wang", "Weiqi Zhang", "Houyu Zhang", "Yufan Guo", "Xian Li", "Bing Yin", "Sujay Sanghavi"]
year: 2026
venue: "Findings of the Association for Computational Linguistics: ACL 2026"
source: "https://aclanthology.org/2026.findings-acl.1344/"
doi: "10.18653/v1/2026.findings-acl.1344"
licence: "CC BY 4.0 (ACL Anthology)"
licence_url: "https://creativecommons.org/licenses/by/4.0/"
summary_licence: "Elamur Cosmic License"
canonical: "https://research.elamur.ai/free/en/papers/zhu-2026-policy-self-distillation-efficient-diffusion/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/zhu-2026-policy-self-distillation-efficient-diffusion/)

[home](https://md.research.elamur.ai/free/en/) / [papers](https://md.research.elamur.ai/free/en/papers/) / [2026](https://md.research.elamur.ai/free/en/papers/2026/page/8/) / On-Policy Self-Distillation for Efficient Diffusion Language Models with Early-Stage Calibration

# On-Policy Self-Distillation for Efficient Diffusion Language Models with Early-Stage Calibration

**Authors:** Huaisheng Zhu, MingYu Liu, Junze Liu, Zhen Ge, [Tian Wang](https://md.research.elamur.ai/free/en/authors/tian-wang/), [Jiri Gesi](https://md.research.elamur.ai/free/en/authors/jiri-gesi/), [Dakuo Wang](https://md.research.elamur.ai/free/en/authors/dakuo-wang/), Weiqi Zhang, [Houyu Zhang](https://md.research.elamur.ai/free/en/authors/houyu-zhang/), [Yufan Guo](https://md.research.elamur.ai/free/en/authors/yufan-guo/), [Xian Li](https://md.research.elamur.ai/free/en/authors/xian-li/), [Bing Yin](https://md.research.elamur.ai/free/en/authors/bing-yin/) and Sujay Sanghavi · **Year:** 2026 · **Venue:** [Findings of the Association for Computational Linguistics: ACL 2026](https://md.research.elamur.ai/free/en/venues/findings/) ([Findings of the Association for Computational Linguistics 2026](https://md.research.elamur.ai/free/en/venues/findings/2026/)) · **Publisher:** [Association for Computational Linguistics](https://md.research.elamur.ai/free/en/publishers/association-for-computational-linguistics/) · **Licence:** [CC BY 4.0 (ACL Anthology)](https://creativecommons.org/licenses/by/4.0/)

**Links:** [Publisher's page](https://aclanthology.org/2026.findings-acl.1344/) · [PDF](https://aclanthology.org/2026.findings-acl.1344.pdf) · [DOI 10.18653/v1/2026.findings-acl.1344](https://doi.org/10.18653/v1/2026.findings-acl.1344) · [OpenAlex](https://openalex.org/W7166881832)

**Citations:** 1 citations (OpenAlex, 2026-10-14) · 1 in our library, of the 149,439 papers whose bibliographies we read

## Summary
Diffusion language models such as masked diffusion models (MDMs) decode many tokens in parallel. Their first decoding steps are unreliable because almost every position is still masked. Those early errors carry forward, lowering output quality. They also force the model to unmask only a few tokens per step, which limits speed.

The authors propose Calibrated On-Policy Self-Distillation (COPSD). It builds an unnormalized target distribution from the model's own predictions at later, better-informed decoding steps. It then distills that target back into the same model through a KL-based objective on self-generated samples. A group relative normalization keeps training stable when few samples are available. No demonstration data or external teacher is required.

Using LLaDA-8B-Instruct, COPSD was evaluated on:
- math (MATH500),
- planning (Sudoku),
- RLHF-style alignment (TL;DR Summarization, Anthropic-HH).

It consistently beats a naive self-distillation baseline and lets the model decode more tokens at once with little quality loss. It also adds gains on top of supervised fine-tuning. This makes it a cheap way to speed up and improve DLLMs.

## Key points
- **Win rates against the reference.** The reference is LLaDA-8B-Instruct decoding 2 tokens per step. At the same 2-token setting, COPSD reaches 52.60% on TL;DR and 64.60% on Anthropic-HH. Naive self-distillation (Base-Self) scores 43.20% and 51.40%.
- **Faster decoding with little loss.** At 4 tokens per step on Anthropic-HH, COPSD still gets a 45.60% win rate, close to parity with the 2-token base model.
- **Combined with SFT.** On TL;DR at 2 tokens, SFT+COPSD reaches 75.40%, compared with 73.00% for SFT alone and 70.20% for SFT+Base-Self.
- **Ablation on MATH500.** With fixed early and late timesteps, COPSD scores 34.60% accuracy. Randomly sampling s/T (COPSD-Random) gives 33.20%, and the base model gives 32.40%.
- **Hyperparameters and limitations.** With t/T fixed at 0.9, values of s/T between 0.5 and 0.7 work best, and sensitivity is low overall. The main limitation is that training depends on online sampling from the current model.

*Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.*

📚 Shelves (4): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 16,667 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 16,194 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 16,039 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 15,891 works

🧭 Topics (2): [🧠 Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) · 3,320 works · [🪶 Efficient models & compression](https://md.research.elamur.ai/free/en/topics/efficient-models/) · 242 works

## Cite

Cite as: Huaisheng Zhu, MingYu Liu, Junze Liu, Zhen Ge, Tian Wang, Jiri Gesi, Dakuo Wang, Weiqi Zhang, Houyu Zhang, Yufan Guo, Xian Li, Bing Yin and Sujay Sanghavi. 2026. On-Policy Self-Distillation for Efficient Diffusion Language Models with Early-Stage Calibration. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work.

```bibtex
@inproceedings{zhu2026policy,
  title = {{On-Policy Self-Distillation for Efficient Diffusion Language Models with Early-Stage Calibration}},
  author = {Huaisheng Zhu, MingYu Liu, Junze Liu, Zhen Ge, Tian Wang, Jiri Gesi, Dakuo Wang, Weiqi Zhang, Houyu Zhang, Yufan Guo, Xian Li, Bing Yin and Sujay Sanghavi},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.1344},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.1344}
}
```

- [← previous On the Mathematical Relationship Between Layer Normalization and Dynamic Activation Functions](https://md.research.elamur.ai/free/en/papers/stollenwerk-2026-mathematical-relationship-between-layer-normalization/)
- [next → Once Correct, Still Wrong: Counterfactual Hallucination in Multilingual Vision-Language Models](https://md.research.elamur.ai/free/en/papers/mousi-2026-once-correct-still-wrong-counterfactual/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/8/)

## References

38 references, 18 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

1. Rishabh Agarwal, Nino Vieillard, Yongchao Zhou, Piotr Stanczyk, Sabela Ramos Garea, Matthieu Geist, and Olivier Bachem. 2024. [On-policy distillation of language models: Learning from self-generated mistakes](https://md.research.elamur.ai/free/en/papers/agarwal-2023-policy-distillation-language-models-learning/). In The twelfth international conference on learning representations.
2. Arel. 2025. Arel’s sudoku generator. [https://www.ocf.berkeley.edu/~arel/sudoku/main.html](https://www.ocf.berkeley.edu/~arel/sudoku/main.html). Accessed: 2025-04-08. [↗](https://www.ocf.berkeley.edu/~arel/sudoku/main.html)
3. Jacob Austin, Daniel D Johnson, Jonathan Ho, Daniel Tarlow, and Rianne Van Den Berg. 2021. [Structured denoising diffusion models in discrete state-spaces](https://md.research.elamur.ai/free/en/papers/austin-2021-structured-denoising-diffusion-models-discrete/). Advances in neural information processing systems, 34:17981–17993.
4. Yuntao Bai, Andy Jones, Kamal Ndousse, Amanda Askell, Anna Chen, Nova DasSarma, Dawn Drain, Stanislav Fort, Deep Ganguli, Tom Henighan, and 1 others. 2022. Training a helpful and harmless assistant with reinforcement learning from human feedback. arXiv preprint arXiv:2204.05862. [find](https://scholar.google.com/scholar?q=Yuntao+Bai%2C+Andy+Jones%2C+Kamal+Ndousse%2C+Amanda+Askell%2C+Anna+Chen%2C+Nova+DasSarma%2C+Dawn+Drain%2C+Stanislav+Fort%2C+Deep+Ganguli%2C+Tom+Henighan%2C+and+1+others%2E+2022%2E+Training+a+helpful+and+harmless+assistant+wi)
5. Andrew Campbell, Joe Benton, Valentin De Bortoli, Thomas Rainforth, George Deligiannidis, and Arnaud Doucet. 2022. A continuous time framework for discrete denoising models. Advances in Neural Information Processing Systems, 35:28266–28279. [find](https://scholar.google.com/scholar?q=Andrew+Campbell%2C+Joe+Benton%2C+Valentin+De+Bortoli%2C+Thomas+Rainforth%2C+George+Deligiannidis%2C+and+Arnaud+Doucet%2E+2022%2E+A+continuous+time+framework+for+discrete+denoising+models%2E+Advances+in+Neural+Informa)
6. Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, and 1 others. 2021. [Training verifiers to solve math word problems](https://md.research.elamur.ai/free/en/papers/cobbe-2021-training-verifiers-solve-math-word/). arXiv preprint arXiv:2110.14168.
7. Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. [Bert: Pre-training of deep bidirectional transformers for language understanding](https://md.research.elamur.ai/free/en/papers/devlin-2019-bert-pre-training-deep-bidirectional/). In Proceedings of the 2019 conference of the North American chapter of the association for computational linguistics: human language technologies,…
8. Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, and 1 others. 2024. [The llama 3 herd of models](https://md.research.elamur.ai/free/en/papers/grattafiori-2024-llama-3-herd-models/). arXiv e-prints, pages arXiv–2407.
9. Shansan Gong, Shivam Agarwal, Yizhe Zhang, Jiacheng Ye, Lin Zheng, Mukai Li, Chenxin An, Peilin Zhao, Wei Bi, Jiawei Han, and 1 others. 2024. Scaling diffusion language models via adaptation from autoregressive models. arXiv preprint arXiv:2410.17891. — [arXiv:2410.17891](https://arxiv.org/abs/2410.17891) [find](https://scholar.google.com/scholar?q=Shansan+Gong%2C+Shivam+Agarwal%2C+Yizhe+Zhang%2C+Jiacheng+Ye%2C+Lin+Zheng%2C+Mukai+Li%2C+Chenxin+An%2C+Peilin+Zhao%2C+Wei+Bi%2C+Jiawei+Han%2C+and+1+others%2E+2024%2E+Scaling+diffusion+language+models+via+adaptation+from+auto)
10. Shansan Gong, Ruixiang Zhang, Huangjie Zheng, Jiatao Gu, Navdeep Jaitly, Lingpeng Kong, and Yizhe Zhang. 2025. Diffucoder: Understanding and improving masked diffusion models for code generation. arXiv preprint arXiv:2506.20639. — [arXiv:2506.20639](https://arxiv.org/abs/2506.20639) [find](https://scholar.google.com/scholar?q=Shansan+Gong%2C+Ruixiang+Zhang%2C+Huangjie+Zheng%2C+Jiatao+Gu%2C+Navdeep+Jaitly%2C+Lingpeng+Kong%2C+and+Yizhe+Zhang%2E+2025%2E+Diffucoder%3A+Understanding+and+improving+masked+diffusion+models+for+code+generation%2E+arXi)
11. Yuxian Gu, Li Dong, Furu Wei, and Minlie Huang. 2023. [Minillm: Knowledge distillation of large language models](https://md.research.elamur.ai/free/en/papers/gu-2023-minillm-policy-distillation-large-language/). arXiv preprint arXiv:2306.08543.
12. Michael Gutmann and Aapo Hyvärinen. 2010. Noise-contrastive estimation: A new estimation principle for unnormalized statistical models. In Proceedings of the thirteenth international conference on artificial intelligence and statistics, pages 297–304. JMLR Workshop and Conference Proceedings. [find](https://scholar.google.com/scholar?q=Michael+Gutmann+and+Aapo+Hyv%C3%A4rinen%2E+2010%2E+Noise%2Dcontrastive+estimation%3A+A+new+estimation+principle+for+unnormalized+statistical+models%2E+In+Proceedings+of+the+thirteenth+international+conference+on+art)
13. Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, and Jacob Steinhardt. 2024. [Measuring mathematical problem solving with the math dataset](https://md.research.elamur.ai/free/en/papers/hendrycks-2021-measuring-mathematical-problem-solving-math/), 2021. URL https://arxiv.org/abs/2103.03874. [↗](https://arxiv.org/abs/2103.03874)
14. Jonas Hübotter, Frederike Lübeck, Lejs Behric, Anton Baumann, Marco Bagatella, Daniel Marta, Ido Hakimi, Idan Shenfeld, Thomas Kleine Buening, Carlos Guestrin, and 1 others. 2026. Reinforcement learning via self-distillation. arXiv preprint arXiv:2601.20802. [find](https://scholar.google.com/scholar?q=Jonas+H%C3%BCbotter%2C+Frederike+L%C3%BCbeck%2C+Lejs+Behric%2C+Anton+Baumann%2C+Marco+Bagatella%2C+Daniel+Marta%2C+Ido+Hakimi%2C+Idan+Shenfeld%2C+Thomas+Kleine+Buening%2C+Carlos+Guestrin%2C+and+1+others%2E+2026%2E+Reinforcement+learni)
15. Jeonghye Kim, Xufang Luo, Minbeom Kim, Sangmook Lee, Dohyung Kim, Jiwon Jeon, Dongsheng Li, and Yuqing Yang. 2026. Why does self-distillation (sometimes) degrade the reasoning capability of llms? arXiv preprint arXiv:2603.24472. — [arXiv:2603.24472](https://arxiv.org/abs/2603.24472) [find](https://scholar.google.com/scholar?q=Jeonghye+Kim%2C+Xufang+Luo%2C+Minbeom+Kim%2C+Sangmook+Lee%2C+Dohyung+Kim%2C+Jiwon+Jeon%2C+Dongsheng+Li%2C+and+Yuqing+Yang%2E+2026%2E+Why+does+self%2Ddistillation+%28sometimes%29+degrade+the+reasoning+capability+of+llms%3F+arXi)
16. Hunter Lightman, Vineet Kosaraju, Yuri Burda, Harrison Edwards, Bowen Baker, Teddy Lee, Jan Leike, John Schulman, Ilya Sutskever, and Karl Cobbe. 2023. [Let’s verify step by step](https://md.research.elamur.ai/free/en/papers/lightman-2023-let-s-verify-step-step/). In The Twelfth International Conference on Learning Representations.
17. Aaron Lou, Chenlin Meng, and Stefano Ermon. 2023. Discrete diffusion language modeling by estimating the ratios of the data distribution. [find](https://scholar.google.com/scholar?q=Aaron+Lou%2C+Chenlin+Meng%2C+and+Stefano+Ermon%2E+2023%2E+Discrete+diffusion+language+modeling+by+estimating+the+ratios+of+the+data+distribution%2E)
18. Chenlin Meng, Kristy Choi, Jiaming Song, and Stefano Ermon. 2022. Concrete score matching: Generalized score matching for discrete data. Advances in Neural Information Processing Systems, 35:34532–34545. [find](https://scholar.google.com/scholar?q=Chenlin+Meng%2C+Kristy+Choi%2C+Jiaming+Song%2C+and+Stefano+Ermon%2E+2022%2E+Concrete+score+matching%3A+Generalized+score+matching+for+discrete+data%2E+Advances+in+Neural+Information+Processing+Systems%2C+35%3A34532%E2%80%93345)
19. Niklas Muennighoff, Zitong Yang, Weijia Shi, Xiang Lisa Li, Li Fei-Fei, Hannaneh Hajishirzi, Luke Zettlemoyer, Percy Liang, Emmanuel Candès, and Tatsunori Hashimoto. 2025. [s1: Simple test-time scaling](https://md.research.elamur.ai/free/en/papers/muennighoff-2025-s1-simple-test-time-scaling/). arXiv preprint arXiv:2501.19393.
20. Shen Nie, Fengqi Zhu, Chao Du, Tianyu Pang, Qian Liu, Guangtao Zeng, Min Lin, and Chongxuan Li. 2024. Scaling up masked diffusion models on text. arXiv preprint arXiv:2410.18514. — [arXiv:2410.18514](https://arxiv.org/abs/2410.18514) [find](https://scholar.google.com/scholar?q=Shen+Nie%2C+Fengqi+Zhu%2C+Chao+Du%2C+Tianyu+Pang%2C+Qian+Liu%2C+Guangtao+Zeng%2C+Min+Lin%2C+and+Chongxuan+Li%2E+2024%2E+Scaling+up+masked+diffusion+models+on+text%2E+arXiv+preprint+arXiv%3A2410%2E18514%2E+%E2%80%94+%5BarXiv%3A2410%2E18514%5D%28)
21. Shen Nie, Fengqi Zhu, Zebin You, Xiaolu Zhang, Jingyang Ou, Jun Hu, Jun Zhou, Yankai Lin, Ji-Rong Wen, and Chongxuan Li. 2025. [Large language diffusion models](https://md.research.elamur.ai/free/en/papers/nie-2025-large-language-diffusion-models/). arXiv preprint arXiv:2502.09992.
22. Jingyang Ou, Shen Nie, Kaiwen Xue, Fengqi Zhu, Jiacheng Sun, Zhenguo Li, and Chongxuan Li. 2024. Your absorbing discrete diffusion secretly models the conditional distributions of clean data. arXiv preprint arXiv:2406.03736. — [arXiv:2406.03736](https://arxiv.org/abs/2406.03736) [find](https://scholar.google.com/scholar?q=Jingyang+Ou%2C+Shen+Nie%2C+Kaiwen+Xue%2C+Fengqi+Zhu%2C+Jiacheng+Sun%2C+Zhenguo+Li%2C+and+Chongxuan+Li%2E+2024%2E+Your+absorbing+discrete+diffusion+secretly+models+the+conditional+distributions+of+clean+data%2E+arXiv+pr)
23. Mihir Prabhudesai, Menging Wu, Amir Zadeh, Katerina Fragkiadaki, and Deepak Pathak. 2025. Diffusion beats autoregressive in data-constrained settings. arXiv preprint arXiv:2507.15857. — [arXiv:2507.15857](https://arxiv.org/abs/2507.15857) [find](https://scholar.google.com/scholar?q=Mihir+Prabhudesai%2C+Menging+Wu%2C+Amir+Zadeh%2C+Katerina+Fragkiadaki%2C+and+Deepak+Pathak%2E+2025%2E+Diffusion+beats+autoregressive+in+data%2Dconstrained+settings%2E+arXiv+preprint+arXiv%3A2507%2E15857%2E+%E2%80%94+%5BarXiv%3A2507%2E15)
24. Subham Sahoo, Marianne Arriola, Yair Schiff, Aaron Gokaslan, Edgar Marroquin, Justin Chiu, Alexander Rush, and Volodymyr Kuleshov. 2024. [Simple and effective masked diffusion language models](https://md.research.elamur.ai/free/en/papers/sahoo-2024-simple-effective-masked-diffusion-language/). Advances in Neural Information Processing Systems, 37:130136–130184.
25. Subham Sekhar Sahoo, Justin Deschenaux, Aaron Gokaslan, Guanghan Wang, Justin Chiu, and Volodymyr Kuleshov. 2025. The diffusion duality. arXiv preprint arXiv:2506.10892. — [arXiv:2506.10892](https://arxiv.org/abs/2506.10892) [find](https://scholar.google.com/scholar?q=Subham+Sekhar+Sahoo%2C+Justin+Deschenaux%2C+Aaron+Gokaslan%2C+Guanghan+Wang%2C+Justin+Chiu%2C+and+Volodymyr+Kuleshov%2E+2025%2E+The+diffusion+duality%2E+arXiv+preprint+arXiv%3A2506%2E10892%2E+%E2%80%94+%5BarXiv%3A2506%2E10892%5D%28https%3A%2F%2Fa)
26. Jiaxin Shi, Kehang Han, Zhe Wang, Arnaud Doucet, and Michalis Titsias. 2024. Simplified and generalized masked diffusion for discrete data. Advances in neural information processing systems, 37:103131– 103167. [find](https://scholar.google.com/scholar?q=Jiaxin+Shi%2C+Kehang+Han%2C+Zhe+Wang%2C+Arnaud+Doucet%2C+and+Michalis+Titsias%2E+2024%2E+Simplified+and+generalized+masked+diffusion+for+discrete+data%2E+Advances+in+neural+information+processing+systems%2C+37%3A103131)
27. Jascha Sohl-Dickstein, Eric Weiss, Niru Maheswaranathan, and Surya Ganguli. 2015. [Deep unsupervised learning using nonequilibrium thermodynamics](https://md.research.elamur.ai/free/en/papers/sohl-dickstein-2015-deep-unsupervised-learning-using-nonequilibrium/). In International conference on machine learning, pages 2256–2265. pmlr.
28. Yang Song and Diederik P Kingma. 2021. How to train your energy-based models. arXiv preprint arXiv:2101.03288. — [arXiv:2101.03288](https://arxiv.org/abs/2101.03288) [find](https://scholar.google.com/scholar?q=Yang+Song+and+Diederik+P+Kingma%2E+2021%2E+How+to+train+your+energy%2Dbased+models%2E+arXiv+preprint+arXiv%3A2101%2E03288%2E+%E2%80%94+%5BarXiv%3A2101%2E03288%5D%28https%3A%2F%2Farxiv%2Eorg%2Fabs%2F2101%2E03288%29)
29. Hugo Touvron, Louis Martin, Kevin Stone, Peter Albert, Amjad Almahairi, Yasmine Babaei, Nikolay Bashlykov, Soumya Batra, Prajjwal Bhargava, Shruti Bhosale, and 1 others. 2023. [Llama 2: Open foundation and fine-tuned chat models](https://md.research.elamur.ai/free/en/papers/touvron-2023-llama-2-open-foundation-fine/). arXiv preprint arXiv:2307.09288.
30. Michael Völske, Martin Potthast, Shahbaz Syed, and Benno Stein. 2017. [Tl; dr: Mining reddit to learn automatic summarization](https://md.research.elamur.ai/free/en/papers/volske-2017-tl-dr-mining-reddit-learn/). In Proceedings of the Workshop on New Frontiers in Summarization, pages 59– 63.
31. Guanghan Wang, Yair Schiff, Subham Sekhar Sahoo, and Volodymyr Kuleshov. 2025. Remasking discrete diffusion models with inference-time scaling. arXiv preprint arXiv:2503.00307. — [arXiv:2503.00307](https://arxiv.org/abs/2503.00307) [find](https://scholar.google.com/scholar?q=Guanghan+Wang%2C+Yair+Schiff%2C+Subham+Sekhar+Sahoo%2C+and+Volodymyr+Kuleshov%2E+2025%2E+Remasking+discrete+diffusion+models+with+inference%2Dtime+scaling%2E+arXiv+preprint+arXiv%3A2503%2E00307%2E+%E2%80%94+%5BarXiv%3A2503%2E00307%5D%28ht)
32. Bangjun Xiao, Bingquan Xia, Bo Yang, Bofei Gao, Bowen Shen, Chen Zhang, Chenhong He, Chiheng Lou, Fuli Luo, Gang Wang, and 1 others. 2026. [Mimo-v2-flash technical report](https://md.research.elamur.ai/free/en/papers/team-2026-llm-mimo-v2-flash-technical/). arXiv preprint arXiv:2601.02780.
33. Wenda Xu, Rujun Han, Zifeng Wang, Long T Le, Dhruv Madeka, Lei Li, William Yang Wang, Rishabh Agarwal, Chen-Yu Lee, and Tomas Pfister. 2024. [Speculative knowledge distillation: Bridging the teacher-student gap through interleaved sampling](https://md.research.elamur.ai/free/en/papers/xu-2024-speculative-knowledge-distillation-bridging-teacher/). arXiv preprint arXiv:2410.11325.
34. Zebin You, Shen Nie, Xiaolu Zhang, Jun Hu, Jun Zhou, Zhiwu Lu, Ji-Rong Wen, and Chongxuan Li. 2025. Llada-v: [Large language diffusion models](https://md.research.elamur.ai/free/en/papers/nie-2025-large-language-diffusion-models/) with visual instruction tuning. arXiv preprint arXiv:2505.16933.
35. Siyan Zhao, Devaansh Gupta, Qinqing Zheng, and Aditya Grover. 2025. d1: Scaling reasoning in diffusion large language models via reinforcement learning. arXiv preprint arXiv:2504.12216. — [arXiv:2504.12216](https://arxiv.org/abs/2504.12216) [find](https://scholar.google.com/scholar?q=Siyan+Zhao%2C+Devaansh+Gupta%2C+Qinqing+Zheng%2C+and+Aditya+Grover%2E+2025%2E+d1%3A+Scaling+reasoning+in+diffusion+large+language+models+via+reinforcement+learning%2E+arXiv+preprint+arXiv%3A2504%2E12216%2E+%E2%80%94+%5BarXiv%3A2504%2E)
36. Siyan Zhao, Zhihui Xie, Mengchen Liu, Jing Huang, Guan Pang, Feiyu Chen, and Aditya Grover. 2026. Self-distilled reasoner: On-policy self-distillation for large language models. arXiv preprint arXiv:2601.18734. [find](https://scholar.google.com/scholar?q=Siyan+Zhao%2C+Zhihui+Xie%2C+Mengchen+Liu%2C+Jing+Huang%2C+Guan+Pang%2C+Feiyu+Chen%2C+and+Aditya+Grover%2E+2026%2E+Self%2Ddistilled+reasoner%3A+On%2Dpolicy+self%2Ddistillation+for+large+language+models%2E+arXiv+preprint+arXiv%3A2)
37. Fengqi Zhu, Rongzhen Wang, Shen Nie, Xiaolu Zhang, Chunwei Wu, Jun Hu, Jun Zhou, Jianfei Chen, Yankai Lin, Ji-Rong Wen, and 1 others. 2025a. [Llada 1.5: Variance-reduced preference optimization for large language diffusion models](https://md.research.elamur.ai/free/en/papers/zhu-2026-llada-1-5-variance-reduced/). arXiv preprint arXiv:2505.19223.
38. Huaisheng Zhu, Zhengyu Chen, Shijie Zhou, Zhihui Xie, Yige Yuan, Shiqi Chen, Zhimeng Guo, Siyuan Xu, Hangfan Zhang, Vasant Honavar, and 1 others. 2025b. Simple denoising diffusion language models. arXiv preprint arXiv:2510.22926. — [arXiv:2510.22926](https://arxiv.org/abs/2510.22926) [find](https://scholar.google.com/scholar?q=Huaisheng+Zhu%2C+Zhengyu+Chen%2C+Shijie+Zhou%2C+Zhihui+Xie%2C+Yige+Yuan%2C+Shiqi+Chen%2C+Zhimeng+Guo%2C+Siyuan+Xu%2C+Hangfan+Zhang%2C+Vasant+Honavar%2C+and+1+others%2E+2025b%2E+Simple+denoising+diffusion+language+models%2E+arX)

*Updated 31 October 2026*