StraighttotheGradient:LearningtoUseNovelTokensforNeuralTextGenerationXiangLin1SimengHan1ShafiqJoty12Abstractrization(Seeetal.,2017),imagecaptioning(Melas-Kyriazietal.,2018;Wang&Chan,2019)andmachine...
HowandWhytoUseExperimentalDatatoEvaluateMethodsforObservationalCausalInferenceAmandaGentzel12PurvaPruthi1DavidJensen1Abstractfectsfromobservationaldata.Suchinterestisunderstand-able,giventhecentral...
CalibrateBeforeUse:ImprovingFew-ShotPerformanceofLanguageModelsTonyZ.Zhao1EricWallace1ShiFeng2DanKlein1SameerSingh3AbstractInput:Subparacting.Sentiment:NegativeGPT-3canperformnumeroustaskswhenpro-I...
CollectatOnce,UseEffectively:MakingNon-interactiveLocallyPrivateLearningPossibleKaiZheng1WenlongMou1LiweiWang1Abstractetal.,2016),machinelearning(Chaudhuri&Monteleoni,2008;Chaudhurietal.,2011;Rubin...