Alignment with human representations supports robust few-shot learning

Open in new window