Short pieces on what happens to communication when a model is in the loop. They are meant to be read in order, oldest first: the question, then the reader, then the manuscripts.
Every summary is a decision about what a reader will never see. Three things worth measuring before we call a generated summary good.
Read →We keep designing tags that say where text came from, and keep being surprised when readers interpret them their own way.
Read →How the three papers in review fit together: what the model writes, how we judge it, and how we read what it produced.
Read →