Writing an outstanding paper on hardware and software co-design is definitely not easy these days. Most of the time, if a paper can define a problem well (no need to be very theoretical or challenging), propose a seemingly logical solution (does not have to be mathematically fancy), provides relevant experimental data (no need for big improvements), make claims judiciously (avoid excessive groundless claims), then it has 95% chance to win the BEST PAPER AWARD at the probability of 0.9.
Reading papers to learn? Lucky you!