In context learning is probably best framed as priming the model towards already learned relations. The model was trained previously. Machine Learning Street T

In context learning is probably best framed as priming the model towards already learned relations. The model was trained previously. Machine Learning Street Talk and Yannik Kilchers coverage of this years Neurips conference give some good broad level overviews of some papers in this regard