Elamur Research⌕
☰
home / papers / 2026 / Breaking Block Boundaries: Anchor-based History-stable Decoding for Diffusion Large Language Models

Breaking Block Boundaries: Anchor-based History-stable Decoding for Diffusion Large Language Models

Paper · 2026 · Shun Zou, Yong Wang, Zehui Chen et al.

Summary

Diffusion LLMs (dLLMs) often use semi-autoregressive (Semi-AR) decoding. This splits the output into blocks that must be finished one after another. The authors show this wastes steps: many tokens in later blocks already hold their final value long before their block is processed. Such early-settled tokens also tend to help neighbouring positions converge. Analysing LLaDA-8B-Instruct, they find three things. Naively decoding ahead on raw confidence is error-prone because confidence fluctuates locally. True stability shows up as a sharp upward confidence trend. Standard decoding ignores earlier iterations entirely. Their remedy is Anchor-based History-stable Decoding (AHD). It needs no training and plugs into existing samplers. At each step, the current prediction acts as an anchor and is checked against a short buffer of past steps. Tokens that stay consistent are released early, even outside the active block. The method is tested on language, vision-language and audio-language models. AHD cuts decoding steps sharply and also raises accuracy, unlike acceleration schemes that usually trade quality for speed.

Key points

  • On BBH, AHD needs about 80% fewer decoding steps and scores 3.67% higher.
  • Long generations (length 1024, LLaDA-8B-Instruct):
    • On BBH, the score rises from 53.28 to 59.82 and the step count drops by about 94%.
    • On HumanEval, the score rises from 43.90 to 48.78 with about 86% fewer steps.
    • AHD beats PC-sampler, Fast-dLLM, KLASS and Saber in this setting.
  • Robustness:
    • Gains hold across block lengths (16–128) and generation lengths on MBPP, MMLU-Pro and BBH.
    • On Mathvista-mini, AHD gives roughly 2.0–2.4× speed-up together with higher scores.
    • Results stay stable across sampling temperatures from 0 to 1 on MMLU-Pro.
  • Ablations: the best trade-off comes with consistency threshold ε = 0.02 and history buffer length H = 6. A very short buffer, such as H = 2, hurts quality noticeably.
  • Limitations: hyperparameters need small per-task tuning, and only models of roughly 8B parameters were tested. Code is available at github.com/zs1314/AHD.

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

 Institutions

🏢Alibaba GroupHangzhou · China201 works · 17,173 citations🎓University of Science and Technology of ChinaHefei · China145 works · 114,409 citations

 Fields, topics and concepts

📚 Shelves 5The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 worksArts and Humanities · 17,450 worksLanguage and Linguistics · 17,296 worksComputer Vision and Pattern Recognition · 3,486 works

🧭 Topics 1Research subjects and languages the work is about.Language models & pretraining · 3,894 works

Cite

Cite as Shun Zou, Yong Wang, Zehui Chen, Lin Chen, Chongyang Tao, Feng Zhao and Xiangxiang Chu. 2026. Breaking Block Boundaries: Anchor-based History-stable Decoding for Diffusion Large Language Models. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-long.945

@inproceedings{zou2026breaking,
  title = {{Breaking Block Boundaries: Anchor-based History-stable Decoding for Diffusion Large Language Models}},
  author = {Shun Zou, Yong Wang, Zehui Chen, Lin Chen, Chongyang Tao, Feng Zhao and Xiangxiang Chu},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-long.945},
  url = {https://doi.org/10.18653/v1/2026.acl-long.945}
}

References

58 references, 20 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Jacob Austin, Daniel D Johnson, Jonathan Ho, Daniel Tarlow, and Rianne Van Den Berg. 2021a. Structured denoising diffusion models in discrete state-spaces. Advances in Neural Information Processing Systems, 34:17981–17993.
  2. Jacob Austin, Augustus Odena, Maxwell Nye, Maarten Bosma, Henryk Michalewski, David Dohan, Ellen Jiang, Carrie Cai, Michael Terry, Quoc Le, and Charles Sutton. 2021b. Program synthesis with large language models. arXiv preprint arXiv:2108.07732.
  3. Heli Ben-Hamu, Itai Gat, Daniel Severo, Niklas Nolte, and Brian Karrer. 2025. Accelerated sampling from masked diffusion models via entropy bounded unmasking. arXiv preprint arXiv:2505.24857. — [arXiv:2505.24857](https://arxiv.org/abs/2505.24857) find
  4. Lukas Berglund, Meg Tong, Max Kaufmann, Mikita Balesni, Asa Cooper Stickland, Tomasz Korbak, and Owain Evans. 2024. The reversal curse: Llms trained on "a is b" fail to learn "b is a". arXiv preprint arXiv:2309.12288. find
  5. Tiwei Bie, Maosong Cao, Kun Chen, Lun Du, Mingliang Gong, Zhuochen Gong, Yanmei Gu, Jiaqi Hu, Zenan Huang, Zhenzhong Lan, Chengxi Li, Chongxuan Li, Jianguo Li, Zehuan Li, Huabin Liu, Lin Liu, Guoshan Lu, Xiaocheng Lu, Yuxin Ma, and 12 others. 2025. Llada2.0: Scaling up diffusion language models to 1… — [arXiv:2512.15745](https://arxiv.org/abs/2512.15745) find
  6. 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
  7. Jiaao Chen, Aston Zhang, Mu Li, Alex Smola, and Diyi Yang. 2023. A cheaper and better diffusion language model with soft-masked noise. In Proceedings of the Conference on Empirical Methods in Natural Language Processing, pages 4765–4775. find
  8. Kecheng Chen, Ziru Liu, Xijia Tao, Hui Liu, Xinyu Fu, Suiyun Zhang, Dandan Tu, Lingpeng Kong, Rui Liu, and Haoliang Li. 2025a. Beyond confidence: Adaptive and coherent decoding for diffusion language models. arXiv preprint arXiv:2512.02044. — [arXiv:2512.02044](https://arxiv.org/abs/2512.02044) find
  9. Mark Chen, Jerry Tworek, Heewoo Jun, Qiming Yuan, Henrique Ponde de Oliveira Pinto, Jared Kaplan, Harri Edwards, Yuri Burda, Nicholas Joseph, Greg Brockman, Alex Ray, Raul Puri, Gretchen Krueger, Michael Petrov, Heidy Khlaaf, Girish Sastry, Pamela Mishkin, Brooke Chan, Scott Gray, and 39 others. 202… — Evaluating Large Language Models Trained on Code
  10. Xinhua Chen, Sitao Huang, Cong Guo, Chiyue Wei, Yintao He, Jianyi Zhang, Hai "Helen" Li, and Yiran Chen. 2025b. Dpad: Efficient diffusion language models with suffix dropout. arXiv preprint arXiv:2508.14148. — [arXiv:2508.14148](https://arxiv.org/abs/2508.14148) find
  11. Yiming Chen, Xianghu Yue, Chen Zhang, Xiaoxue Gao, Robby T. Tan, and Haizhou Li. 2024. Voicebench: Benchmarking llm-based voice assistants. arXiv preprint arXiv:2410.17196.
  12. Karl Cobbe, Vineet Kosaraju, Mohammad Bavarian, Mark Chen, Heewoo Jun, Lukasz Kaiser, Matthias Plappert, Jerry Tworek, Jacob Hilton, Reiichiro Nakano, Christopher Hesse, and John Schulman. 2021. Training verifiers to solve math word problems. arXiv preprint arXiv:2110.14168.
  13. Yihong Dong, Zhaoyu Ma, Xue Jiang, Zhiyuan Fan, Jiaru Qian, Yongmin Li, Jianha Xiao, Zhi Jin, Rongyu Cao, Binhua Li, Fei Huang, Yongbin Li, and Ge Li. 2025. Saber: An efficient sampling with adaptive acceleration and backtracking enhanced remasking for diffusion language model. arXiv preprint arXiv:… — [arXiv:2510.18165](https://arxiv.org/abs/2510.18165) find
  14. Chaoyou Fu, Peixian Chen, Yunhang Shen, Yulei Qin, Mengdan Zhang, Xu Lin, Jinrui Yang, Xiawu Zheng, Ke Li, Xing Sun, Yunsheng Wu, Rongrong Ji, Caifeng Shan, and Ran He. 2025. Mme: A comprehensive evaluation benchmark for multimodal large language models. arXiv preprint arXiv:2306.13394.
  15. Shansan Gong, Shivam Agarwal, Yizhe Zhang, Jiacheng Ye, Lin Zheng, Mukai Li, Chenxin An, Peilin Zhao, Wei Bi, Jiawei Han, Hao Peng, and Lingpeng Kong. 2025. Scaling diffusion language models via adaptation from autoregressive models. arXiv preprint arXiv:2410.17891. — [arXiv:2410.17891](https://arxiv.org/abs/2410.17891) find
  16. Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, and Jacob Steinhardt. 2021. Measuring mathematical problem solving with the math dataset. arXiv preprint arXiv:2103.03874.
  17. Jonathan Ho, Ajay Jain, and Pieter Abbeel. 2020. Denoising diffusion probabilistic models. Advances in Neural Information Processing Systems, 33:6840– 6851. find
  18. Feng Hong, Geng Yu, Yushi Ye, Haicheng Huang, Huangjie Zheng, Ya Zhang, Yanfeng Wang, and Jiangchao Yao. 2025. Wide-in, narrow-out: Revok-able decoding for efficient and effective dllms. arXiv preprint arXiv:2507.18578. — [arXiv:2507.18578](https://arxiv.org/abs/2507.18578) find
  19. Jianuo Huang, Yaojie Zhang, Yicun Yang, Benhao Huang, Biqing Qi, Dongrui Liu, and Linfeng Zhang. 2025a. Mask tokens as prophet: Fine-grained cache eviction for efficient dllm inference. arXiv preprint arXiv:2510.09309. find
  20. Pengcheng Huang, Shuhao Liu, Zhenghao Liu, Yukun Yan, Shuo Wang, Zulong Chen, and Tong Xiao. 2025b. Pc-sampler: Position-aware calibration of decoding bias in masked diffusion models. arXiv preprint arXiv:2508.13021. — [arXiv:2508.13021](https://arxiv.org/abs/2508.13021) find
  21. Stephanie Lin, Jacob Hilton, and Owain Evans. 2022. Truthfulqa: Measuring how models mimic human falsehoods. In Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 3214–3252.
  22. Zhiyuan Liu, Yicun Yang, Yaojie Zhang, Junjie Chen, Chang Zou, Qingyuan Wei, Shaobo Wang, and Linfeng Zhang. 2025. dllm-cache: Accelerating diffusion large language models with adaptive caching. arXiv preprint arXiv:2506.06295. — [arXiv:2506.06295](https://arxiv.org/abs/2506.06295) find
  23. Justin Lovelace, Varsha Kishore, Chao Wan, Eliot Shekhtman, and Kilian Q Weinberger. 2023. Latent diffusion for language generation. Advances in Neural Information Processing Systems, 36:56998– 57025. find
  24. Guanxi Lu, Hao Mark Chen, Yuto Karashima, Zhi-can Wang, Daichi Fujiki, and Hongxiang Fan. 2026. Adablock-dllm: Semantic-aware diffusion llm inference via adaptive block size. arXiv preprint arXiv:2509.26432. — [arXiv:2509.26432](https://arxiv.org/abs/2509.26432) find
  25. Pan Lu, Hritik Bansal, Tony Xia, Jiacheng Liu, Chunyuan Li, Hannaneh Hajishirzi, Hao Cheng, Kai-Wei Chang, Michel Galley, and Jianfeng Gao. 2024. Mathvista: Evaluating mathematical reasoning of foundation models in visual contexts. arXiv preprint arXiv:2310.02255.
  26. Pan Lu, Swaroop Mishra, Tony Xia, Liang Qiu, Kai-Wei Chang, Song-Chun Zhu, Oyvind Tafjord, Peter Clark, and Ashwin Kalyan. 2022. Learn to explain: Multimodal reasoning via thought chains for science question answering. Advances in Neural Information Processing Systems, 35:2507–2521. find
  27. Xinyin Ma, Runpeng Yu, Gongfan Fang, and Xinchao Wang. 2025a. dkv-cache: The cache for diffusion language models. arXiv preprint arXiv:2505.15781. — [arXiv:2505.15781](https://arxiv.org/abs/2505.15781) find
  28. Drew A Hudson and Christopher D Manning. 2019. Gqa: A new dataset for real-world visual reasoning and compositional question answering. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 6700–6709.
  29. Yuxin Ma, Lun Du, Lanning Wei, Kun Chen, Qian Xu, Kangyu Wang, Guofeng Feng, Guoshan Lu, Lin Liu, Xiaojing Qi, Xinyuan Zhang, Zhen Tao, Haibo Feng, Ziyun Jiang, Ying Xu, Zenan Huang, Yihong Zhuang, Haokai Xu, Jiaqi Hu, and 4 others. 2025b. dinfer: An efficient inference framework for diffusion langu… find
  30. Seo Hyun Kim, Sunwoo Hong, Hojung Jung, Youngrok Park, and Se-Young Yun. 2026. Klass: Kl-guided fast inference in masked diffusion models. arXiv preprint arXiv:2511.05664. — [arXiv:2511.05664](https://arxiv.org/abs/2511.05664) find
  31. Andrew Melnik, Michal Ljubljanac, Cong Lu, Qi Yan, Weiming Ren, and Helge Ritter. 2024. Video diffusion models: A survey. arXiv preprint arXiv:2405.03150. — [arXiv:2405.03150](https://arxiv.org/abs/2405.03150) find
  32. Fanheng Kong, Jingyuan Zhang, Yahui Liu, Zirui Wu, Yu Tian, Victoria W., and Guorui Zhou. 2025. Accelerating diffusion llm inference via local determinism propagation. arXiv preprint arXiv:2510.07081. — [arXiv:2510.07081](https://arxiv.org/abs/2510.07081) find
  33. Shen-Yun Miao, Chao-Chun Liang, and Keh-Yih Su. 2020. A diverse corpus for evaluating and developing english math word problem solvers. In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics, pages 975–984. find
  34. Pengxiang Li, Yefan Zhou, Dilxat Muhtar, Lu Yin, Shilin Yan, Li Shen, Soroush Vosoughi, and Shiwei Liu. 2026. Diffusion language models know the answer before decoding. arXiv preprint arXiv:2508.19982. — [arXiv:2508.19982](https://arxiv.org/abs/2508.19982) find
  35. 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. arXiv preprint arXiv:2502.09992.
  36. Jingyang Ou, Shen Nie, Kaiwen Xue, Fengqi Zhu, Jiacheng Sun, Zhenguo Li, and Chongxuan Li. 2026. 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
  37. Ku, Kai Wang, Alex Zhuang, Rongqi Fan, Xiang Yue, and Wenhu Chen. 2024b. Mmlu-pro: A more robust and challenging multi-task language understanding benchmark. Advances in Neural Information Processing Systems, 37:95266–95290.
  38. Robin Rombach, Andreas Blattmann, Dominik Lorenz, Patrick Esser, and Björn Ommer. 2022. High-resolution image synthesis with latent diffusion models. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 10684–10695.
  39. Qingyan Wei, Yaojie Zhang, Zhiyuan Liu, Puyu Zeng, Yuxuan Wang, Biqing Qi, Dongrui Liu, and Linfeng Zhang. 2026. Accelerating diffusion large language models with slowfast sampling: The three golden principles. arXiv preprint arXiv:2506.10848. — [arXiv:2506.10848](https://arxiv.org/abs/2506.10848) find
  40. Subham Sekhar Sahoo, Marianne Arriola, Yair Schiff, Aaron Gokaslan, Edgar Marroquin, Justin T Chiu, Alexander Rush, and Volodymyr Kuleshov. 2024. Simple and effective masked diffusion language models. Advances in Neural Information Processing Systems, 37:130136–130184.
  41. Jiaxin Shi, Kehang Han, Zhe Wang, Arnaud Doucet, and Michalis K. Titsias. 2024. Simplified and generalized masked diffusion for discrete data. Advances in Neural Information Processing Systems, 37:103131– 103167.
  42. Yuerong Song, Xiaoran Liu, Ruixiao Li, Zhigeng Liu, Zengfeng Huang, Qipeng Guo, Ziwei He, and Xipeng Qiu. 2025. Sparse-dllm: Accelerating diffusion llms with dynamic cache eviction. arXiv preprint arXiv:2508.02558. — [arXiv:2508.02558](https://arxiv.org/abs/2508.02558) find
  43. Mirac Suzgun, Nathan Scales, Nathanael Schärli, Sebastian Gehrmann, Yi Tay, Hyung Won Chung, Aakanksha Chowdhery, Quoc V. Le, Ed H. Chi, Denny Zhou, and Jason Wei. 2023. Challenging BIG-bench tasks and whether chain-of-thought can solve them. In Findings of the Association for Computational Linguist…
  44. Kangyu Wang, Zhiyun Jiang, Haibo Feng, Weijia Zhao, Lin Liu, Jianguo Li, Zhenzhong Lan, and Weiyao Lin. 2025a. Creditdecoding: Accelerating parallel decoding in diffusion large language models with trace credits. arXiv preprint arXiv:2510.06133. — [arXiv:2510.06133](https://arxiv.org/abs/2510.06133) find
  45. Ke Wang, Junting Pan, Weikang Shi, Zimu Lu, Houxing Ren, Aojun Zhou, Mingjie Zhan, and Hongsheng Li. 2024a. Measuring multimodal mathematical reasoning with math-vision dataset. Advances in Neural Information Processing Systems, 37:95095–95169.
  46. Wen Wang, Bozhen Fang, Chenchen Jing, Yongliang Shen, Yangyi Shen, Qiuyu Wang, Hao Ouyang, Hao Chen, and Chunhua Shen. 2025b. Time is a feature: Exploiting temporal dynamics in diffusion language models. arXiv preprint arXiv:2508.09138. — [arXiv:2508.09138](https://arxiv.org/abs/2508.09138) find
  47. Xu Wang, Chenkai Xu, Yijie Jin, Jiachun Jin, Hao Zhang, and Zhijie Deng. 2025c. Diffusion llms can do faster-than-ar inference via discrete diffusion forcing. arXiv preprint arXiv:2508.09192. — [arXiv:2508.09192](https://arxiv.org/abs/2508.09192) find
  48. Yubo Wang, Xueguang Ma, Ge Zhang, Yuansheng Ni, Abhranil Chandra, Shiguang Guo, Weiming Ren, Aaran Arulraj, Xuan He, Ziyan Jiang, Tianle Li, Max Chengyue Wu, Hao Zhang, Shuchen Xue, Zhijian Liu, Shizhe Diao, Ligeng Zhu, Ping Luo, Song Han, and Enze Xie. 2025. Fast-dllm: Training-free acceleration of… — [arXiv:2505.22618](https://arxiv.org/abs/2505.22618) find
  49. Yi Xin, Qi Qin, Siqi Luo, Kaiwen Zhu, Juncheng Yan, Yan Tai, Jiayi Lei, Yuewen Cao, Keqi Wang, Yibin Wang, Jinbin Bai, Qian Yu, Dengyang Jiang, Yuandong Pu, Haoxing Chen, Le Zhuo, Junjun He, Gen Luo, Tianbin Li, and 13 others. 2025. Lumina-dimoo: An omni diffusion large language model for multimodal… — [arXiv:2510.06308](https://arxiv.org/abs/2510.06308) find
  50. Ling Yang, Ye Tian, Bowen Li, Xinchen Zhang, Ke Shen, Yunhai Tong, and Mengdi Wang. 2025. Mmada: Multimodal large diffusion language models. arXiv preprint arXiv:2505.15809. — [arXiv:2505.15809](https://arxiv.org/abs/2505.15809) find
  51. Ling Yang, Zhilong Zhang, Yang Song, Shenda Hong, Runsheng Xu, Yue Zhao, Wentao Zhang, Bin Cui, and Ming-Hsuan Yang. 2023. Diffusion models: A comprehensive survey of methods and applications. ACM Computing Surveys, 56(4):1–39. find
  52. Jiacheng Ye, Zhihui Xie, Lin Zheng, Jiahui Gao, Zirui Wu, Xin Jiang, Zhenguo Li, and Lingpeng Kong. 2025. Dream 7b: Diffusion large language models. arXiv preprint arXiv:2508.15487.
  53. Runpeng Yu, Qi Li, and Xinchao Wang. 2025a. Discrete diffusion in large language and multimodal models: A survey. arXiv preprint arXiv:2506.13759. — [arXiv:2506.13759](https://arxiv.org/abs/2506.13759) find
  54. Runpeng Yu, Xinyin Ma, and Xinchao Wang. 2025b. Dimple: Discrete diffusion multimodal large language model with parallel decoding. arXiv preprint arXiv:2505.16990. — [arXiv:2505.16990](https://arxiv.org/abs/2505.16990) find
  55. Kaichen Zhang, Bo Li, Peiyuan Zhang, Fanyi Pu, Joshua Adrian Cahyono, Kairui Hu, Shuai Liu, Yuanhan Zhang, Jingkang Yang, Chunyuan Li, and Ziwei Liu. 2025. LMMs-eval: Reality check on the evaluation of large multimodal models. In Findings of the Association for Computational Linguistics: NAACL 2025,…
  56. Tianyi Zhang, Zheng Wang, Jing Huang, Mohiuddin Muhammad Tasnim, and Wei Shi. 2023. A survey of diffusion based image generation models: Issues and their solutions. arXiv preprint arXiv:2308.13142. — [arXiv:2308.13142](https://arxiv.org/abs/2308.13142) find
  57. Jiaming Zhou, Hongjie Chen, Shiwan Zhao, Jian Kang, Jie Li, Enzhi Wang, Yujie Guo, Haoqin Sun, Hui Wang, Aobo Kong, Yong Qin, and Xuelong Li. 2025. Diffa: Large language diffusion models can listen and understand. arXiv preprint arXiv:2507.18452. — [arXiv:2507.18452](https://arxiv.org/abs/2507.18452) find
  58. Fengqi Zhu, Rongzhen Wang, Shen Nie, Xiaolu Zhang, Chunwei Wu, Jun Hu, Jun Zhou, Jianfei Chen, Yankai Lin, Ji-Rong Wen, and Chongxuan Li. 2025. Llada 1.5: Variance-reduced preference optimization for large language diffusion models. arXiv preprint arXiv:2505.19223.

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zou-2026-breaking-block-boundaries-anchor-based/