{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,16]],"date-time":"2026-07-16T12:04:59Z","timestamp":1784203499424,"version":"3.55.0"},"reference-count":36,"publisher":"Elsevier BV","license":[{"start":{"date-parts":[[2026,10,1]],"date-time":"2026-10-01T00:00:00Z","timestamp":1790812800000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/www.elsevier.com\/tdm\/userlicense\/1.0\/"},{"start":{"date-parts":[[2026,10,1]],"date-time":"2026-10-01T00:00:00Z","timestamp":1790812800000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/www.elsevier.com\/legal\/tdmrep-license"},{"start":{"date-parts":[[2026,6,22]],"date-time":"2026-06-22T00:00:00Z","timestamp":1782086400000},"content-version":"vor","delay-in-days":0,"URL":"http:\/\/creativecommons.org\/licenses\/by\/4.0\/"}],"funder":[{"DOI":"10.13039\/100018919","name":"Pengcheng Laboratory","doi-asserted-by":"publisher","award":["PCL2023AS1-5"],"award-info":[{"award-number":["PCL2023AS1-5"]}],"id":[{"id":"10.13039\/100018919","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/100018919","name":"Pengcheng Laboratory","doi-asserted-by":"publisher","award":["PCL2025A09"],"award-info":[{"award-number":["PCL2025A09"]}],"id":[{"id":"10.13039\/100018919","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["elsevier.com","sciencedirect.com"],"crossmark-restriction":true},"short-container-title":["Neural Networks"],"published-print":{"date-parts":[[2026,10]]},"DOI":"10.1016\/j.neunet.2026.109014","type":"journal-article","created":{"date-parts":[[2026,4,25]],"date-time":"2026-04-25T06:12:53Z","timestamp":1777097573000},"page":"109014","update-policy":"https:\/\/doi.org\/10.1016\/elsevier_cm_policy","source":"Crossref","is-referenced-by-count":0,"special_numbering":"C","title":["Optimizing speculative decoding via dynamic multi-path and dual-stream networks"],"prefix":"10.1016","volume":"202","author":[{"ORCID":"https:\/\/orcid.org\/0000-0002-3309-4643","authenticated-orcid":false,"given":"Yong","family":"Yang","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ting Ting","family":"Yang","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shao Shuai","family":"Gao","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jia Hong","family":"Ning","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ling Zheng","family":"Kong","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"78","reference":[{"key":"10.1016\/j.neunet.2026.109014_bib0001","doi-asserted-by":"crossref","unstructured":"Bae, S., Ko, J., Song, H., & Yun, S.-Y. (2023). Fast and robust early-exiting framework for autoregressive language models with synchronized parallel decoding. [Online]. Available: arXiv: 2310.05424.","DOI":"10.18653\/v1\/2023.emnlp-main.362"},{"key":"10.1016\/j.neunet.2026.109014_bib0002","unstructured":"Bai, Y., Lv, X., Zhang, J., Lyu, H., & Tang, J., et al. (2024). Longbench: A bilingual, multitask benchmark for long context understanding. [Online]. Available: arXiv: 2308.14508."},{"key":"10.1016\/j.neunet.2026.109014_bib0003","unstructured":"BELLEGroup (2023). Belle: Be everyone\u2019s large language model engine. https:\/\/github.com\/LianjiaTech\/BELLE."},{"key":"10.1016\/j.neunet.2026.109014_bib0004","unstructured":"Cai, T., Li, Y., Geng, Z., Peng, H., Lee, J. D., Chen, D., & Dao, T. (2024). Medusa: Simple LLM inference acceleration framework with multiple decoding heads. [Online]. Available: arXiv: 2401.10774."},{"key":"10.1016\/j.neunet.2026.109014_bib0005","unstructured":"Chen, C., Borgeaud, S., Irving, G., Lespiau, J.-B., Sifre, L., & Jumper, J. (2023). Accelerating large language model decoding with speculative sampling. [Online]. Available: arXiv: 2302.01318."},{"key":"10.1016\/j.neunet.2026.109014_bib0006","unstructured":"Chiang, W.-L., Li, Z., Lin, Z., Sheng, Y., Wu, Z., Zhang, H., Zheng, L., Zhuang, S., Zhuang, Y., Gonzalez, J. E., Stoica, I., & Xing, E. P. (2023). Vicuna: An open-source chatbot impressing GPT-4 with 90%* chatGPT quality. [Online]. Available: https:\/\/lmsys.org\/blog\/2023-03-30-vicuna\/."},{"key":"10.1016\/j.neunet.2026.109014_bib0007","unstructured":"Chowdhery, A., Narang, S., Devlin, J., Bosma, M., & Mishra, G., et al. (2022). PaLM: Scaling language modeling with pathways. [Online]. Available: arXiv: 2204.02311."},{"key":"10.1016\/j.neunet.2026.109014_bib0008","doi-asserted-by":"crossref","unstructured":"Dettmers, T., Lewis, M., Belkada, Y., & Zettlemoyer, L. (2022). Llm.int8: 8-bit matrix multiplication for transformers at scale. [Online]. Available: arXiv: 2208.07339.","DOI":"10.52202\/068431-2198"},{"key":"10.1016\/j.neunet.2026.109014_bib0009","unstructured":"Dubey, A., Jauhri, A., Pandey, A., Kadian, A., & Al-Dahle, A., et al. (2024). The llama 3 herd of models. [Online]. Available: arXiv: 2407.21783."},{"key":"10.1016\/j.neunet.2026.109014_bib0010","unstructured":"Elbayad, M., Gu, J., Grave, E., & Auli, M. (2020). Depth-adaptive transformer. [Online]. Available: arXiv: 1910.10073."},{"key":"10.1016\/j.neunet.2026.109014_bib0011","unstructured":"Fu, Y., Bailis, P., Stoica, I., & Zhang, H. (2024). Break the sequential dependency of LLM inference using lookahead decoding. [Online]. Available: arXiv: 2402.02057."},{"key":"10.1016\/j.neunet.2026.109014_bib0012","unstructured":"Hoffmann, J., Borgeaud, S., Mensch, A., Buchatskaya, E., & Cai, T., et al. (2022). Training compute-optimal large language models. [Online]. Available: arXiv: 2203.15556."},{"key":"10.1016\/j.neunet.2026.109014_bib0013","doi-asserted-by":"crossref","unstructured":"Kim, S., Mangalam, K., Moon, S., Malik, J., & Mahoney, M. W., et al. (2023). Speculative decoding with big little decoder. [Online]. Available: arXiv: 2302.07863.","DOI":"10.52202\/075280-1705"},{"key":"10.1016\/j.neunet.2026.109014_bib0014","unstructured":"Kim, T., Kim, J., Lee, G., & Yun, S.-Y. (2024). Instructive decoding: Instruction-tuned large language models are self-refiner from noisy instructions. arXiv: 2311.00233."},{"key":"10.1016\/j.neunet.2026.109014_bib0015","unstructured":"Leviathan, Y., Kalman, M., & Matias, Y. (2023). Fast inference from transformers via speculative decoding. [Online]. Available: arXiv: 2211.17192."},{"key":"10.1016\/j.neunet.2026.109014_bib0016","unstructured":"Li, Y., Wei, F., Zhang, C., & Zhang, H. (2024). Eagle: Speculative sampling requires rethinking feature uncertainty. [Online]. Available: arXiv: 2401.15077."},{"key":"10.1016\/j.neunet.2026.109014_bib0017","unstructured":"Li, Z., Kovachki, N., Azizzadenesheli, K., Liu, B., & Bhattacharya, K., et al. (2021). Fourier neural operator for parametric partial differential equations. [Online]. Available: arXiv: 2010.08895."},{"key":"10.1016\/j.neunet.2026.109014_bib0018","unstructured":"Lu, X., Zeng, Y., Ma, F., Yu, Z., & Levorato, M. (2024). Improving multi-candidate speculative decoding. [Online]. Available: arXiv: 2409.10644."},{"key":"10.1016\/j.neunet.2026.109014_bib0019","doi-asserted-by":"crossref","unstructured":"Ma, X., Fang, G., & Wang, X. (2023). Llm-pruner: On the structural pruning of large language models. [Online]. Available: arXiv: 2305.11627.","DOI":"10.52202\/075280-0950"},{"key":"10.1016\/j.neunet.2026.109014_bib0020","unstructured":"Meta (2024). Llama3. https:\/\/github.com\/pytorch-labs\/gpt-fast\/."},{"key":"10.1016\/j.neunet.2026.109014_bib0021","series-title":"Proceedings of the 29th ACM international conference on architectural support for programming languages and operating systems, volume 3","article-title":"Specinfer: Accelerating large language model serving with tree-based speculative inference and verification","author":"Miao","year":"2024"},{"key":"10.1016\/j.neunet.2026.109014_bib0022","unstructured":"Monea, G., Joulin, A., & Grave, E. (2023). PaSS: Parallel speculative sampling. [Online]. Available: arXiv: 2311.13581."},{"key":"10.1016\/j.neunet.2026.109014_bib0023","doi-asserted-by":"crossref","unstructured":"Quan, G., Feng, W., Hao, C., Jiang, G., Zhang, Y., & Wang, H. (2025). Rasd: Retrieval-augmented speculative decoding. [Online]. Available: arXiv: 2503.03434.","DOI":"10.18653\/v1\/2025.findings-acl.320"},{"key":"10.1016\/j.neunet.2026.109014_bib0024","doi-asserted-by":"crossref","unstructured":"Schuster, T., Fisch, A., Gupta, J., Dehghani, M., & Bahri, D., et al. (2022). Confident adaptive language modeling. [Online]. Available: arXiv: 2207.07061.","DOI":"10.52202\/068431-1269"},{"key":"10.1016\/j.neunet.2026.109014_bib0025","unstructured":"Stern, M., Shazeer, N., & Uszkoreit, J. (2018). Blockwise parallel decoding for deep autoregressive models. arXiv: 1811.03115."},{"key":"10.1016\/j.neunet.2026.109014_bib0026","unstructured":"Sun, M., Liu, Z., Bair, A., & Kolter, J. Z. (2024). A simple and effective pruning approach for large language models. [Online]. Available: arXiv: 2306.11695."},{"key":"10.1016\/j.neunet.2026.109014_bib0027","unstructured":"Taori, R., Gulrajani, I., Zhang, T., Dubois, Y., & Li, X., et al. (2023). Stanford alpaca: An instruction-following LLaMA model. https:\/\/github.com\/tatsu-lab\/stanford_alpaca."},{"key":"10.1016\/j.neunet.2026.109014_bib0028","unstructured":"G. Team, Mesnard, T., Hardin, C., Dadashi, R., & Bhupatiraju, S., et al. (2024). Gemma: Open models based on gemini research and technology. [Online]. Available: arXiv: 2403.08295."},{"key":"10.1016\/j.neunet.2026.109014_bib0029","unstructured":"Touvron, H., Lavril, T., Izacard, G., Martinet, X., & Lachaux, M.-A., et al. (2023). Llama: Open and efficient foundation language models. [Online]. Available: arXiv: 2302.13971."},{"key":"10.1016\/j.neunet.2026.109014_bib0030","unstructured":"Wang, Z., Azmat, M., Li, A., Horesh, R., & Yurochkin, M. (2025). Speculate, then collaborate: Fusing knowledge of language models during decoding. [Online]. Available: arXiv: 2502.08020."},{"key":"10.1016\/j.neunet.2026.109014_bib0031","unstructured":"Wiggers, & Kyle (2024). OpenAI debuts GPT-4o \u2019omni\u2019 model now powering chatGPT. [Online]. Available:TechCrunch. https:\/\/techcrunch.com\/openais-newest-model-is-gpt-4o\/."},{"key":"10.1016\/j.neunet.2026.109014_bib0032","doi-asserted-by":"crossref","unstructured":"Xia, H., Yang, Z., Dong, Q., Wang, P., Li, Y., & Ge, T., et al. (2024). Unlocking efficiency in large language model inference: A comprehensive survey of speculative decoding. [Online]. Available: arXiv: 2401.07851.","DOI":"10.18653\/v1\/2024.findings-acl.456"},{"key":"10.1016\/j.neunet.2026.109014_bib0033","unstructured":"Xiao, G., Lin, J., Seznec, M., Wu, H., Demouth, J., & Han, S. (2024). Smoothquant: Accurate and efficient post-training quantization for large language models. [Online]. Available: arXiv: 2211.10438."},{"key":"10.1016\/j.neunet.2026.109014_bib0034","unstructured":"Yang, N., Ge, T., Wang, L., Jiao, B., & Jiang, D., et al. (2023). Inference with reference: Lossless acceleration of large language models. [Online]. Available: arXiv: 2304.04487."},{"key":"10.1016\/j.neunet.2026.109014_bib0035","unstructured":"Zhang, S., Roller, S., Goyal, N., Artetxe, M., Chen, M., Chen, S., Dewan, C., Diab, M., Li, X., Lin, X. V., Mihaylov, T., Ott, M., Shleifer, S., Shuster, K., Simig, D., Koura, P. S., Sridhar, A., Wang, T., & Zettlemoyer, L. (2022). Opt: Open pre-trained transformer language models. [Online]. Available: arXiv: 2205.01068."},{"key":"10.1016\/j.neunet.2026.109014_bib0036","doi-asserted-by":"crossref","unstructured":"Zheng, L., Chiang, W.-L., Sheng, Y., Zhuang, S., & Wu, Z., et al. (2023). Judging LLM-as-a-judge with MT-bench and chatbot arena.","DOI":"10.52202\/075280-2020"}],"container-title":["Neural Networks"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/api.elsevier.com\/content\/article\/PII:S0893608026004740?httpAccept=text\/xml","content-type":"text\/xml","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/api.elsevier.com\/content\/article\/PII:S0893608026004740?httpAccept=text\/plain","content-type":"text\/plain","content-version":"vor","intended-application":"text-mining"}],"deposited":{"date-parts":[[2026,7,16]],"date-time":"2026-07-16T11:16:00Z","timestamp":1784200560000},"score":1,"resource":{"primary":{"URL":"https:\/\/linkinghub.elsevier.com\/retrieve\/pii\/S0893608026004740"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,10]]},"references-count":36,"alternative-id":["S0893608026004740"],"URL":"https:\/\/doi.org\/10.1016\/j.neunet.2026.109014","relation":{},"ISSN":["0893-6080"],"issn-type":[{"value":"0893-6080","type":"print"}],"subject":[],"published":{"date-parts":[[2026,10]]},"assertion":[{"value":"Elsevier","name":"publisher","label":"This article is maintained by"},{"value":"Optimizing speculative decoding via dynamic multi-path and dual-stream networks","name":"articletitle","label":"Article Title"},{"value":"Neural Networks","name":"journaltitle","label":"Journal Title"},{"value":"https:\/\/doi.org\/10.1016\/j.neunet.2026.109014","name":"articlelink","label":"CrossRef DOI link to publisher maintained version"},{"value":"article","name":"content_type","label":"Content Type"},{"value":"\u00a9 2026 The Authors. Published by Elsevier Ltd.","name":"copyright","label":"Copyright"}],"article-number":"109014"}}