From the paper's caption for that figure:

Regional trends of extreme temperatures are underestimated in climate model experiments in multiple regions globally. (A) Comparison of observed trends in tail-widening (yearly 99th percentile minus 87.5th percentile) with 49 simulations from coupled and SST-forced climate models (SI Appendix, Table S1). Observed trends are outside of the modeled range in several regions globally (dark red). Areas where the annual 87.5th percentile of Tx shows a negative trend in observations are shown in gray ~~ So I think that means tail-widening ~= more heat waves. Colored areas are where heat waves are increasing, and climate models have underestimated the observed heat waves. Gray = areas that heat waves have become less frequent~~ nevermind, I'm still confused lol

The caption from the columbia article seems wrong:

Regions where observed heat waves exceed trends from climate models. Boxed areas with the darkest red colors are the most extreme; lesser reds and oranges exceed models, but not by as much. Yellows roughly match models, while greens and blues are below what models would project.

Pretty sure that's wrong because the scale on the paper's graphic shows a scale going from blue (0%) to dark red (100%). That means the map is much worse than what article says. Yellow is not where it "roughly match models", it's actually where the models underestimate by 50-75%