You can see this in the figure below, which shows how global warming in CMIP6 models (red lines) compare to historical temperature observations (blue). Using our emulator that is calibrated to CMIP6 results, we can produce a range of projections (light grey). However, when we introduce our observational constraints, we substantially narrow the range of uncertainty in future projections (dark grey) by eliminating some of the more implausible projections, producing a best estimate future projection (black) that closely follows the observed warming. This method corrects for some of the systematic biases in CMIP6 models, for example a tendency to underpredict the warming in the late 20th Century.