Research papers that probably should exist given the graph of knowledge that it has. I spent about 15 minutes searching for a non-existent paper (I knew this was probably the case when I started, by choice, by choice) and it turns out the title was a mashup of two other papers in the domain.

Same error a human might make if they read too many papers too quickly. Or went to a film festival and was asked to describe the plot(s) of the movies they watched.