EUVIMEDEUVIMED entdecken
Crossref

HyperInfer: Accelerating Large Language Model Inference via Importance-Informed Prefix-KV Caching and Prefetching

Ruidong Zhang, Weijian Chen, Ping Chen, Shuibing He, Wenjie Liang, Rongchen Liu, Siling Yang, Xuechen Zhang

IEEE Transactions on Computers · 2026-12 · 75 (12) · 4091-4105

DOI: 10.1109/tc.2026.3714652 ↗

Kurz gefasst

Fachartikel in der Zeitschrift IEEE Transactions on Computers, 2026-12. Verfasst von Ruidong Zhang und 7 weiteren Autor:innen. Die Arbeit nennt 45 Literaturangaben. Gefördert von: National Natural Science Foundation of China, Major Projects of Zhejiang Province, Hangzhou Chengxi Sci-Tech Innovation Corridor Special Development Fund.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Kein Abstract verfügbar.

Förderung

  • National Natural Science Foundation of China
  • Major Projects of Zhejiang Province
  • Hangzhou Chengxi Sci-Tech Innovation Corridor Special Development Fund
  • Key Research Project of Zhejiang Lab
  • Science and Technology Program of Zhejiang Province

Zitieren

APA

Zhang, R., Chen, W., Chen, P., He, S., Liang, W., Liu, R., Yang, S., & Zhang, X. (2026). HyperInfer: Accelerating Large Language Model Inference via Importance-Informed Prefix-KV Caching and Prefetching. IEEE Transactions on Computers, 75(12), 4091-4105. https://doi.org/10.1109/tc.2026.3714652

Vancouver

Zhang R, Chen W, Chen P, He S, Liang W, Liu R, et al. HyperInfer: Accelerating Large Language Model Inference via Importance-Informed Prefix-KV Caching and Prefetching. IEEE Transactions on Computers. 2026;75(12):4091-4105. doi:10.1109/tc.2026.3714652

BibTeX
@article{Zhang2026,
  author = {Zhang, Ruidong and Chen, Weijian and Chen, Ping and He, Shuibing and Liang, Wenjie and Liu, Rongchen and Yang, Siling and Zhang, Xuechen},
  title = {HyperInfer: Accelerating Large Language Model Inference via Importance-Informed Prefix-KV Caching and Prefetching},
  journal = {IEEE Transactions on Computers},
  year = {2026},
  volume = {75},
  number = {12},
  pages = {4091--4105},
  doi = {10.1109/tc.2026.3714652}
}
Kostenlos starten →