lol... Train your model with the dataset of "another list" i.e. populate a lookup table. Slap a sigmoid on top of that boolean "presence" result. Perfect precision/recall!