The cost of the tokens used to output a typical academic paper length document is on the order of cents, even for frontier models. The vast majority of the cost is in the discussion leading up to that final output - there's no reason someone would try to produce a shorter paper just to save tokens.