This is yet another paper that finds quickly-diminishing returns of reflective thinking.And this paper's result may be more severe: Figure 3 (above) shows accuracy on medical question-answering improved on second thought, but either plateaued or DECLINED *after* second thought.