Gemini：一个高性能多模态模型（英文版）.docx

下载文档

0
0
约15.82万字
约 99页
2024-12-28 发布于山西
举报
版权申诉
保障服务

Gemini：一个高性能多模态模型（英文版）.docx

1、本文档共99页，可阅读全部内容。
2、有哪些信誉好的足球投注网站（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。
3、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击这里二次下载。
4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。

GoogleDeepMind

?2023Google.Allrightsreserved

Gemini:AFamilyofHighlyCapableMultimodalModels

GeminiTeam,Google1

Thisreportintroducesanewfamilyofmultimodalmodels,Gemini,thatexhibitremarkablecapabilitiesacrossimage,audio,video,andtextunderstanding.TheGeminifamilyconsistsofUltra,Pro,andNanosizes,suitableforapplicationsrangingfromcomplexreasoningtaskstoon-devicememory-constraineduse-cases.Evaluationonabroadrangeofbenchmarksshowsthatourmost-capableGeminiUltramodeladvancesthestateoftheartin30of32ofthesebenchmarks—notablybeingthefirstmodeltoachievehuman-expertperformanceonthewell-studiedexambenchmarkMMLU,andimprovingthestateoftheartineveryoneofthe20multimodalbenchmarksweexamined.WebelievethatthenewcapabilitiesofGeminimodelsincross-modalreasoningandlanguageunderstandingwillenableawidevarietyofusecasesandwediscussourapproachtowarddeployingthemresponsiblytousers.

1.Introduction

WepresentGemini,afamilyofhighlycapablemultimodalmodelsdevelopedatGoogle.WetrainedGeminijointlyacrossimage,audio,video,andtextdataforthepurposeofbuildingamodelwithbothstronggeneralistcapabilitiesacrossmodalitiesalongsidecutting-edgeunderstandingandreasoningperformanceineachrespectivedomain.

Gemini1.0,ourfirstversion,comesinthreesizes:Ultraforhighly-complextasks,Proforenhancedperformanceanddeployabilityatscale,andNanoforon-deviceapplications.Eachsizeisspecificallytailoredtoaddressdifferentcomputationallimitationsandapplicationrequirements.WeevaluatetheperformanceofGeminimodelsonacomprehensivesuiteofinternalandexternalbenchmarkscoveringawiderangeoflanguage,coding,reasoning,andmultimodaltasks.

Geminiadvancesstate-of-the-artinlarge-scalelanguagemodeling(Aniletal.,2023;Brownetal.,2020;Chowdheryetal.,2023;Hoffmanneta