Stronger Random Baselines for In-Context Learning

Gregory Yauney, David Mimno

Evaluating the in-context learning classification performance of language models poses challenges due to small dataset sizes, extensive prompt-selection using the validation set, a…