---
title: "SpecBound — paper by Wen and Feng (2026)"
description: "Self-draft speculative decoding uses the base LLM's own shallow layers to propose tokens. 40 references."
authors: ["Zhuofan Wen", "Yang Feng"]
year: 2026
venue: "Findings of the Association for Computational Linguistics: ACL 2026"
source: "https://aclanthology.org/2026.findings-acl.802/"
doi: "10.18653/v1/2026.findings-acl.802"
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/wen-2026-specbound-adaptive-bounded-self-speculation/"
---
[Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/papers/wen-2026-specbound-adaptive-bounded-self-speculation/)

[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/50/) / SpecBound: Adaptive Bounded Self-Speculation with Layer-wise Confidence Calibration

# SpecBound: Adaptive Bounded Self-Speculation with Layer-wise Confidence Calibration

**Authors:** [Zhuofan Wen](https://md.research.elamur.ai/free/en/authors/zhuofan-wen/) and [Yang Feng](https://md.research.elamur.ai/free/en/authors/yang-feng/) · **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.802/) · [PDF](https://aclanthology.org/2026.findings-acl.802.pdf) · [DOI 10.18653/v1/2026.findings-acl.802](https://doi.org/10.18653/v1/2026.findings-acl.802) · [OpenAlex](https://openalex.org/W7166853010)

**Citations:** 0 citations (OpenAlex, 2026-10-14)

## Summary
Self-draft speculative decoding uses the base LLM's own shallow layers to propose tokens. This avoids training a separate draft model, but early-exit variants have so far given modest speedups. The authors probed every layer on an MT-Bench example and found two causes. First, shallow layers often assign high confidence to wrong tokens, because pretraining supervises only the final layer. Second, a few hard tokens push whole draft batches through deep layers, which wastes compute.

SpecBound addresses both problems:
- **Annealed Confidence Threshold:** applies a temperature that decreases with depth, so shallow-layer distributions are flattened before a fixed exit threshold is checked.
- **Bounded Speculation with Cached States:** ends a drafting round when a token fails to exit within the depth limit or the draft reaches its maximum length. It then sends the cached hidden states of all draft tokens through the remaining layers in one parallel pass.

The outputs are identical to the original model's, and the base weights stay frozen. The method reaches up to 2.33× wall-time speedup across long-form tasks and several architectures, which makes lossless acceleration easier to deploy.

## Key points
- Temperature schedule: T_ℓ = 1 + α(1 − ℓ/L). It is highest at layer 1 and falls to 1 at the last layer. Exit happens at the first layer where top-1 probability ≥ τ.
- The method is not fully training-free: it uses lightweight intermediate-layer LM heads trained following AdaDecode. Its hyperparameters also require empirical tuning.
- Latency tests used Vicuna-7B on SpecBench with NVIDIA H800 GPUs, max length 1024 and batch size 1. Per-token latency and throughput improved, while TTFT stayed nearly equal to the baseline.
- Memory overhead is a small fixed cache table (L × w_max) that is reused for every draft window, so long contexts do not inflate it. The method also works beyond greedy decoding.
- Code: github.com/ictnlp/SpecBound (ICT/CAS).

*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 (5): [🖥️ Computer Science](https://md.research.elamur.ai/free/en/shelves/computer-science/) · 18,614 works · [🤖 Artificial Intelligence](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · 17,965 works · [🎭 Arts and Humanities](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · 17,450 works · [🗣️ Language and Linguistics](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · 17,296 works · [👁️ Computer Vision and Pattern Recognition](https://md.research.elamur.ai/free/en/shelves/computer-science/computer-vision-and-pattern-recognition/) · 3,486 works

## Cite

Cite as: Zhuofan Wen and Yang Feng. 2026. SpecBound: Adaptive Bounded Self-Speculation with Layer-wise Confidence 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{wen2026specbound,
  title = {{SpecBound: Adaptive Bounded Self-Speculation with Layer-wise Confidence Calibration}},
  author = {Zhuofan Wen and Yang Feng},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.802},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.802}
}
```

- [← previous Speaking of Language: Reflections on Metalanguage Research in NLP](https://md.research.elamur.ai/free/en/papers/schneider-2026-speaking-language-reflections-metalanguage-research/)
- [next → SpecCache: Speculative KV Cache Reuse for Efficient RAG Serving](https://md.research.elamur.ai/free/en/papers/wen-2026-speccache-speculative-kv-cache-reuse/)
- [↑ back to papers of 2026](https://md.research.elamur.ai/free/en/papers/2026/page/50/)

## References

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

1. Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, and 1 others. 2023. [Gpt-4 technical report](https://md.research.elamur.ai/free/en/papers/openai-2023-gpt-4-technical-report/). arXiv preprint arXiv:2303.08774.
2. Tolga Bolukbasi, Joseph Wang, Ofer Dekel, and Venkatesh Saligrama. 2017. Adaptive neural networks for efficient inference. In International conference on machine learning, pages 527–536. PMLR. [find](https://scholar.google.com/scholar?q=Tolga+Bolukbasi%2C+Joseph+Wang%2C+Ofer+Dekel%2C+and+Venkatesh+Saligrama%2E+2017%2E+Adaptive+neural+networks+for+efficient+inference%2E+In+International+conference+on+machine+learning%2C+pages+527%E2%80%93536%2E+PMLR%2E)
3. Mengyu Bu and Yang Feng. 2026. Language on demand, knowledge at core: Composing llms with encoder-decoder translation models for extensible multilinguality. arXiv preprint arXiv:2603.17512. [find](https://scholar.google.com/scholar?q=Mengyu+Bu+and+Yang+Feng%2E+2026%2E+Language+on+demand%2C+knowledge+at+core%3A+Composing+llms+with+encoder%2Ddecoder+translation+models+for+extensible+multilinguality%2E+arXiv+preprint+arXiv%3A2603%2E17512%2E)
4. Tianle Cai, Yuhong Li, Zhengyang Geng, Hongwu Peng, and Tri Dao. 2023. Medusa: Simple framework for accelerating llm generation with multiple decoding heads. [find](https://scholar.google.com/scholar?q=Tianle+Cai%2C+Yuhong+Li%2C+Zhengyang+Geng%2C+Hongwu+Peng%2C+and+Tri+Dao%2E+2023%2E+Medusa%3A+Simple+framework+for+accelerating+llm+generation+with+multiple+decoding+heads%2E)
5. Charlie Chen, Sebastian Borgeaud, Geoffrey Irving, Jean-Baptiste Lespiau, Laurent Sifre, and John Jumper. 2023. Accelerating large language model decoding with speculative sampling. arXiv preprint arXiv:2302.01318. [find](https://scholar.google.com/scholar?q=Charlie+Chen%2C+Sebastian+Borgeaud%2C+Geoffrey+Irving%2C+Jean%2DBaptiste+Lespiau%2C+Laurent+Sifre%2C+and+John+Jumper%2E+2023%2E+Accelerating+large+language+model+decoding+with+speculative+sampling%2E+arXiv+preprint+arX)
6. Wei-Lin Chiang, Zhuohan Li, Zi Lin, Ying Sheng, Zhanghao Wu, Hao Zhang, Lianmin Zheng, Siyuan Zhuang, Yonghao Zhuang, Joseph E Gonzalez, and 1 others. 2023. Vicuna: An open-source chatbot impressing gpt-4 with 90%* chatgpt quality. See [https://vicuna.lmsys](https://vicuna.lmsys). org (accessed 14 April 2023), 2(3):6. [↗](https://vicuna.lmsys)
7. Yung-Sung Chuang, Yujia Xie, Hongyin Luo, Yoon Kim, James Glass, and Pengcheng He. 2023. [Dola: Decoding by contrasting layers improves factuality in large language models](https://md.research.elamur.ai/free/en/papers/chuang-2023-dola-decoding-contrasting-layers-improves/). arXiv preprint arXiv:2309.03883.
8. Mostafa Elhoushi, Akshat Shrivastava, Diana Liskovich, Basil Hosmer, Bram Wasti, Liangzhen Lai, Anas Mahmoud, Bilge Acun, Saurabh Agarwal, Ahmed Roman, and 1 others. 2024. [Layerskip: Enabling early exit inference and self-speculative decoding](https://md.research.elamur.ai/free/en/papers/elhoushi-2024-layerskip-enabling-early-exit-inference/). In Proceedings of the 62nd Annual Meeting of the Associa…
9. Yichao Fu, Peter Bailis, Ion Stoica, and Hao Zhang. 2024. [Break the sequential dependency of llm inference using lookahead decoding](https://md.research.elamur.ai/free/en/papers/fu-2024-break-sequential-dependency-llm-inference/). arXiv preprint arXiv:2402.02057.
10. Xinwei Geng, Xiaocheng Feng, and Bing Qin. 2021. Learning to rewrite for non-autoregressive neural machine translation. In Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 3297–3308. [find](https://scholar.google.com/scholar?q=Xinwei+Geng%2C+Xiaocheng+Feng%2C+and+Bing+Qin%2E+2021%2E+Learning+to+rewrite+for+non%2Dautoregressive+neural+machine+translation%2E+In+Proceedings+of+the+2021+Conference+on+Empirical+Methods+in+Natural+Language+P)
11. Fabian Gloeckle, Badr Youbi Idrissi, Baptiste Rozière, David Lopez-Paz, and Gabriel Synnaeve. 2024. [Better & faster large language models via multi-token prediction](https://md.research.elamur.ai/free/en/papers/gloeckle-2024-better-faster-large-language-models/). arXiv preprint arXiv:2404.19737.
12. Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, and 1 others. 2025. [Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning](https://md.research.elamur.ai/free/en/papers/deepseek-ai-2025-deepseek-r1-incentivizing-reasoning-capability/). arXiv preprint arXiv:2501.12948.
13. Zhenyu He, Zexuan Zhong, Tianle Cai, Jason Lee, and Di He. 2024. [Rest: Retrieval-based speculative decoding](https://md.research.elamur.ai/free/en/papers/he-2023-rest-retrieval-based-speculative-decoding/). In Proceedings of the 2024 conference of the North American chapter of the association for computational linguistics: Human language technologies (volume 1: long papers), pages 1582–1595.
14. Bo-Kyeong Kim, Geonmin Kim, Tae-Ho Kim, Thibault Castells, Shinkook Choi, Junho Shin, and Hyoung-Kyu Song. 2024. Shortened llama: A simple depth pruning for large language models. arXiv preprint arXiv:2402.02834, 11:1. — [arXiv:2402.02834](https://arxiv.org/abs/2402.02834) [find](https://scholar.google.com/scholar?q=Bo%2DKyeong+Kim%2C+Geonmin+Kim%2C+Tae%2DHo+Kim%2C+Thibault+Castells%2C+Shinkook+Choi%2C+Junho+Shin%2C+and+Hyoung%2DKyu+Song%2E+2024%2E+Shortened+llama%3A+A+simple+depth+pruning+for+large+language+models%2E+arXiv+preprint+arXiv)
15. Jun Kong, Jin Wang, Liang-Chih Yu, and Xuejie Zhang. 2022. Accelerating inference for pretrained language models by unified multi-perspective early exiting. In Proceedings of the 29th International Conference on Computational Linguistics, pages 4677–4686. [find](https://scholar.google.com/scholar?q=Jun+Kong%2C+Jin+Wang%2C+Liang%2DChih+Yu%2C+and+Xuejie+Zhang%2E+2022%2E+Accelerating+inference+for+pretrained+language+models+by+unified+multi%2Dperspective+early+exiting%2E+In+Proceedings+of+the+29th+International+Co)
16. Yaniv Leviathan, Matan Kalman, and Yossi Matias. 2023. Fast inference from transformers via speculative decoding. In International Conference on Machine Learning, pages 19274–19286. PMLR. [find](https://scholar.google.com/scholar?q=Yaniv+Leviathan%2C+Matan+Kalman%2C+and+Yossi+Matias%2E+2023%2E+Fast+inference+from+transformers+via+speculative+decoding%2E+In+International+Conference+on+Machine+Learning%2C+pages+19274%E2%80%9319286%2E+PMLR%2E)
17. Xiaonan Li, Yunfan Shao, Tianxiang Sun, Hang Yan, Xipeng Qiu, and Xuan-Jing Huang. 2021. Accelerating bert inference for sequence labeling via early-exit. In Proceedings of the 59th annual meeting of the Association for Computational Linguistics and the 11th international joint conference on natural… [find](https://scholar.google.com/scholar?q=Xiaonan+Li%2C+Yunfan+Shao%2C+Tianxiang+Sun%2C+Hang+Yan%2C+Xipeng+Qiu%2C+and+Xuan%2DJing+Huang%2E+2021%2E+Accelerating+bert+inference+for+sequence+labeling+via+early%2Dexit%2E+In+Proceedings+of+the+59th+annual+meeting+of+)
18. Yuhui Li, Fangyun Wei, Chao Zhang, and Hongyang Zhang. 2024. [Eagle: Speculative sampling requires rethinking feature uncertainty](https://md.research.elamur.ai/free/en/papers/li-2024-eagle-speculative-sampling-requires-rethinking/). arXiv preprint arXiv:2401.15077.
19. Feng Lin, Hanling Yi, Yifan Yang, Hongbin Li, Xiaotian Yu, Guangming Lu, and Rong Xiao. 2025. Bita: Bidirectional tuning for lossless acceleration in large language models. Expert Systems with Applications, 279:127305. [find](https://scholar.google.com/scholar?q=Feng+Lin%2C+Hanling+Yi%2C+Yifan+Yang%2C+Hongbin+Li%2C+Xiaotian+Yu%2C+Guangming+Lu%2C+and+Rong+Xiao%2E+2025%2E+Bita%3A+Bidirectional+tuning+for+lossless+acceleration+in+large+language+models%2E+Expert+Systems+with+Applica)
20. Fangcheng Liu, Yehui Tang, Zhenhua Liu, Yunsheng Ni, Kai Han, and Yunhe Wang. 2024. Kangaroo: Lossless self-speculative decoding via double early exiting. arXiv preprint arXiv:2404.18911. — [arXiv:2404.18911](https://arxiv.org/abs/2404.18911) [find](https://scholar.google.com/scholar?q=Fangcheng+Liu%2C+Yehui+Tang%2C+Zhenhua+Liu%2C+Yunsheng+Ni%2C+Kai+Han%2C+and+Yunhe+Wang%2E+2024%2E+Kangaroo%3A+Lossless+self%2Dspeculative+decoding+via+double+early+exiting%2E+arXiv+preprint+arXiv%3A2404%2E18911%2E+%E2%80%94+%5BarXiv%3A240)
21. Xin Men, Mingyu Xu, Qingyu Zhang, Qianhao Yuan, Bingning Wang, Hongyu Lin, Yaojie Lu, Xianpei Han, and Weipeng Chen. 2025. [Shortgpt: Layers in large language models are more redundant than you expect](https://md.research.elamur.ai/free/en/papers/men-2024-shortgpt-layers-large-language-models/). In Findings of the Association for Computational Linguistics: ACL 2025, pages 20192–20204.
22. 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.
23. Baptiste Roziere, Jonas Gehring, Fabian Gloeckle, Sten Sootla, Itai Gat, Xiaoqing Ellen Tan, Yossi Adi, Jingyu Liu, Romain Sauvestre, Tal Remez, and 1 others. 2023. [Code llama: Open foundation models for code](https://md.research.elamur.ai/free/en/papers/roziere-2024-code-llama-open-foundation-models/). arXiv preprint arXiv:2308.12950.
24. Yongliang Shen, Kaitao Song, Xu Tan, Dongsheng Li, Weiming Lu, and Yueting Zhuang. 2023. [Hugginggpt: Solving ai tasks with chatgpt and its friends in hugging face](https://md.research.elamur.ai/free/en/papers/shen-2023-hugginggpt-solving-ai-tasks-chatgpt/). Advances in Neural Information Processing Systems, 36:38154–38180.
25. Mitchell Stern, Noam Shazeer, and Jakob Uszkoreit. 2018. Blockwise parallel decoding for deep autoregressive models. Advances in Neural Information Processing Systems, 31. [find](https://scholar.google.com/scholar?q=Mitchell+Stern%2C+Noam+Shazeer%2C+and+Jakob+Uszkoreit%2E+2018%2E+Blockwise+parallel+decoding+for+deep+autoregressive+models%2E+Advances+in+Neural+Information+Processing+Systems%2C+31%2E)
26. Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, and 1 others. 2023. [Llama: Open and efficient foundation language models](https://md.research.elamur.ai/free/en/papers/touvron-2023-llama-open-efficient-foundation-language/). arXiv preprint arXiv:2302.13971.
27. Neeraj Varshney, Agneet Chatterjee, Mihir Parmar, and Chitta Baral. 2023. Accelerating llm inference by enabling intermediate layer decoding. CoRR. [find](https://scholar.google.com/scholar?q=Neeraj+Varshney%2C+Agneet+Chatterjee%2C+Mihir+Parmar%2C+and+Chitta+Baral%2E+2023%2E+Accelerating+llm+inference+by+enabling+intermediate+layer+decoding%2E+CoRR%2E)
28. Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin. 2017. [Attention is all you need](https://md.research.elamur.ai/free/en/papers/vaswani-2017-attention-all-need/). Advances in neural information processing systems, 30.
29. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, and 1 others. 2022. [Chain-of-thought prompting elicits reasoning in large language models](https://md.research.elamur.ai/free/en/papers/wei-2022-chain-thought-prompting-elicits-reasoning/). Advances in neural information processing systems, 35:24824– 24837.
30. Zhepei Wei, Wei-Lin Chen, Xinyu Zhu, and Yu Meng. 2025. Adadecode: Accelerating llm decoding with adaptive layer parallelism. arXiv preprint arXiv:2506.03700. seq2seq generation. In Findings of the Association for Computational Linguistics: EMNLP 2023, pages 3909–3925. — [arXiv:2506.03700](https://arxiv.org/abs/2506.03700) [find](https://scholar.google.com/scholar?q=Zhepei+Wei%2C+Wei%2DLin+Chen%2C+Xinyu+Zhu%2C+and+Yu+Meng%2E+2025%2E+Adadecode%3A+Accelerating+llm+decoding+with+adaptive+layer+parallelism%2E+arXiv+preprint+arXiv%3A2506%2E03700%2E+seq2seq+generation%2E+In+Findings+of+the+As)
31. Heming Xia, Chak Tou Leong, Wenjie Wang, Yongqi Li, and Wenjie Li. 2025. [Tokenskip: Controllable chain-of-thought compression in llms](https://md.research.elamur.ai/free/en/papers/xia-2025-tokenskip-controllable-chain-thought-compression/). arXiv preprint arXiv:2502.12067.
32. Heming Xia, Zhe Yang, Qingxiu Dong, Peiyi Wang, Yongqi Li, Tao Ge, Tianyu Liu, Wenjie Li, and Zhifang Sui. 2024. [Unlocking efficiency in large language model inference: A comprehensive survey of speculative decoding](https://md.research.elamur.ai/free/en/papers/xia-2024-unlocking-efficiency-large-language-model/). arXiv preprint arXiv:2401.07851.
33. Yifei Yang, Zouying Cao, and Hai Zhao. 2024. Laco: Large language model pruning via layer collapse. arXiv preprint arXiv:2402.11187. [find](https://scholar.google.com/scholar?q=Yifei+Yang%2C+Zouying+Cao%2C+and+Hai+Zhao%2E+2024%2E+Laco%3A+Large+language+model+pruning+via+layer+collapse%2E+arXiv+preprint+arXiv%3A2402%2E11187%2E)
34. Hanling Yi, Feng Lin, Hongbin Li, Ning Peiyang, Xiaotian Yu, and Rong Xiao. 2024. Generation meets verification: Accelerating large language model inference with smart parallel auto-correct decoding. In Findings of the Association for Computational Linguistics: ACL 2024, pages 5285–5299. [find](https://scholar.google.com/scholar?q=Hanling+Yi%2C+Feng+Lin%2C+Hongbin+Li%2C+Ning+Peiyang%2C+Xiaotian+Yu%2C+and+Rong+Xiao%2E+2024%2E+Generation+meets+verification%3A+Accelerating+large+language+model+inference+with+smart+parallel+auto%2Dcorrect+decoding%2E+)
35. Dewen Zeng, Nan Du, Tao Wang, Yuanzhong Xu, Tao Lei, Zhifeng Chen, and Claire Cui. 2023. Learning to skip for language modeling. arXiv preprint arXiv:2311.15436. [find](https://scholar.google.com/scholar?q=Dewen+Zeng%2C+Nan+Du%2C+Tao+Wang%2C+Yuanzhong+Xu%2C+Tao+Lei%2C+Zhifeng+Chen%2C+and+Claire+Cui%2E+2023%2E+Learning+to+skip+for+language+modeling%2E+arXiv+preprint+arXiv%3A2311%2E15436%2E)
36. Jun Zhang, Jue Wang, Huan Li, Lidan Shou, Ke Chen, Gang Chen, and Sharad Mehrotra. 2024. [Draft& verify: Lossless large language model acceleration via self-speculative decoding](https://md.research.elamur.ai/free/en/papers/zhang-2024-draft-verify-lossless-large-language/). In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages…
37. Lingzhe Zhang, Liancheng Fang, Chiming Duan, Minghua He, Leyi Pan, Pei Xiao, Shiyu Huang, Yunpeng Zhai, Xuming Hu, Philip S Yu, and 1 others. 2025. A survey on parallel text generation: From parallel decoding to diffusion language models. arXiv preprint arXiv:2508.08712. — [arXiv:2508.08712](https://arxiv.org/abs/2508.08712) [find](https://scholar.google.com/scholar?q=Lingzhe+Zhang%2C+Liancheng+Fang%2C+Chiming+Duan%2C+Minghua+He%2C+Leyi+Pan%2C+Pei+Xiao%2C+Shiyu+Huang%2C+Yunpeng+Zhai%2C+Xuming+Hu%2C+Philip+S+Yu%2C+and+1+others%2E+2025%2E+A+survey+on+parallel+text+generation%3A+From+parallel+)
38. Jingze Zhu, Yongliang Wu, Wenbo Zhu, Jiawang Cao, Yanqiang Zheng, Jiawei Chen, Xu Yang, Bernt Schiele, Jonas Fischer, and Xinting Hu. 2025. Layercake: Token-aware contrastive decoding within large language model layers. arXiv preprint arXiv:2507.04404. — [arXiv:2507.04404](https://arxiv.org/abs/2507.04404) [find](https://scholar.google.com/scholar?q=Jingze+Zhu%2C+Yongliang+Wu%2C+Wenbo+Zhu%2C+Jiawang+Cao%2C+Yanqiang+Zheng%2C+Jiawei+Chen%2C+Xu+Yang%2C+Bernt+Schiele%2C+Jonas+Fischer%2C+and+Xinting+Hu%2E+2025%2E+Layercake%3A+Token%2Daware+contrastive+decoding+within+large+lan)
39. Zhuofan Wen, Shangtong Gui, and Yang Feng. 2024. Speculative decoding with ctc-based draft model for llm inference acceleration. Advances in Neural Information Processing Systems, 37:92082–92100. [find](https://scholar.google.com/scholar?q=Zhuofan+Wen%2C+Shangtong+Gui%2C+and+Yang+Feng%2E+2024%2E+Speculative+decoding+with+ctc%2Dbased+draft+model+for+llm+inference+acceleration%2E+Advances+in+Neural+Information+Processing+Systems%2C+37%3A92082%E2%80%9392100%2E)
40. Heming Xia, Tao Ge, Peiyi Wang, Si-Qing Chen, Furu Wei, and Zhifang Sui. 2023. Speculative decoding: Exploiting speculative execution for accelerating [find](https://scholar.google.com/scholar?q=Heming+Xia%2C+Tao+Ge%2C+Peiyi+Wang%2C+Si%2DQing+Chen%2C+Furu+Wei%2C+and+Zhifang+Sui%2E+2023%2E+Speculative+decoding%3A+Exploiting+speculative+execution+for+accelerating)

*Updated 2 November 2026*