Idanwo Tuning Prefix

Awọn idahun ati awọn alaye han bi o ṣe n mu ibeere naa.

1. Kini iṣatunṣe iṣatunṣe gangan ṣe ikẹkọ?
   • Gbogbo awọn ti awọn awoṣe ká transformer òṣuwọn
   • Nikan awọn awoṣe ká o wu fokabulari Layer
   • Ẹda lọtọ ti awoṣe fun iṣẹ-ṣiṣe kọọkan
   • Eto kekere ti awọn olutọpa ìpele iṣaaju nigba ti awoṣe naa duro ni aotoju
2. Nibo ni a ti itasi awọn abẹrẹ ìpele?
   • Sinu awọn bọtini ati awọn iye ni gbogbo akiyesi Layer
   • Nikan ni ik o wu Layer
   • Sinu iṣẹ isonu nikan
   • Sinu tokenizer ṣaaju ki o to fi sii
3. Tani o ṣe afihan iṣatunṣe iṣatunṣe ati aijọju nigbawo?
   • OpenAI ni ọdun 2018
   • Li ati Liang ni Stanford ni ọdun 2021
   • Google ni ọdun 2015
   • Meta ni ọdun 2023
4. Kini idi ti iṣatunṣe iṣaaju jẹ wuni fun ṣiṣe ọpọlọpọ awọn iṣẹ ṣiṣe?
   • Ko nilo data ikẹkọ rara
   • O mu ki inference ṣiṣẹ lori ko si hardware
   • Iṣẹ-ṣiṣe kọọkan nilo asọtẹlẹ kekere kan dipo ẹda awoṣe kikun
   • O tun ṣe atunṣe awoṣe ipilẹ fun gbogbo awọn iṣẹ-ṣiṣe ni ẹẹkan
5. Ẹtan wo ni igbagbogbo lo lati ṣe iduroṣinṣin ikẹkọ iṣaaju?
   • Ilọpo meji oṣuwọn ikẹkọ ni gbogbo igbesẹ
   • Didi ìpele ati ikẹkọ awoṣe
   • Yiyọ gbogbo akiyesi fẹlẹfẹlẹ
   • Atunse awọn ìpele nipasẹ kan kekere kikọ sii siwaju nẹtiwọki ti o ti wa ni nigbamii sọnu
6. Lori iru iṣẹ wo ni iṣatunṣe iṣaju iṣaju paapaa tan imọlẹ?
   • Iran ede adayeba bi tabili-si-ọrọ ati akopọ
   • Pipa Pipa Pipa
   • Audio igbi kolaginni
   • Atọka aaye data
7. What proportion of model parameters did the original prefix-tuning paper report learning in its experiments?
   • About 10%
   • About 0.1%
   • Nipa 50%
   • All model parameters
8. How did prefix-tuning perform in the paper’s low-data settings?
   • It always matched full fine-tuning on every task
   • It could not train with limited examples
   • It outperformed full fine-tuning in some evaluated low-data cases
   • It changed every backbone parameter