Tweeted By @ak92501
True Few-Shot Learning with Language Models
— AK (@ak92501) May 25, 2021
pdf: https://t.co/a0WvisxlJX
abs: https://t.co/7zb714iSYk
github: https://t.co/nPICjpoRuo
few-shot ability of LMs when held-out examples are unavailable, prior work significantly overestimated the true few-shot ability of LMs pic.twitter.com/k28CCi4iWk