DeepSeek模型关键创新技术综述.docxVIP

  • 4
  • 0
  • 约3.77万字
  • 约 11页
  • 2026-09-30 发布于江苏
  • 举报

PAGE

PAGE10

AReviewofDeepSeekModels’KeyInnovativeTechniques

ChengenWangR

UniversityofTexasatDallas

chengen.wang@

MuratKantarciogluVirginiaTech

muratk@

arXiv:2503.11486v1[cs.LG]14Mar2025

arXiv:2503.11486v1[cs.LG]14Mar2025

Abstract

DeepSeek-V3andDeepSeek-R1areleadingopen-sourceLargeLanguageModels(LLMs)forgeneral-purposetasksandreasoning,achievingperformancecomparabletostate-of-the-artclosed-sourcemodelsfromcompanieslikeOpenAIandAnthropic—whilerequiringonlyafractionoftheirtrainingcosts.Understandingthekeyinnovativetechnique

您可能关注的文档

文档评论(0)

1亿VIP精品文档

相关文档