Where one study disagrees with what other studies report for the same models, and what we are doing to resolve each case.
These are automatic leads, not findings of error. Each release, a check compares how every study ranks the models it shares with other studies. It flags a study that reverses the order the others agree on, or whose gap to a shared reference model lies well outside theirs. Different tasks, populations or metrics often explain a flag.
A flag never hides a result. Flagged results stay in every table, page and download, marked Contested. A result is withheld only when a reviewer marks it disputed.
Brandes et al. 2023 Supplementary Table 2: full benchmark evaluations of ESM1b and all other methods ranks SIFT above PROVEAN in 3 of 3 comparisons; 3 of 3 other studies rank PROVEAN above SIFT.
5 papers rank PROVEAN above SIFT in 32 of their 39 comparisons and SIFT higher in 4; 3 are ties. Brandes et al. 2023 ranks SIFT above PROVEAN in all 3 comparisons that are not ties (1 is a tie).
ProteinGym: Large-Scale Benchmarks for Protein Fitness Prediction and Design ranks ESM3 above ESM-2 in 18 of 18 comparisons; pfmbench primary benchmark evidence ranks ESM-2 above ESM3 in 9 of 12 comparisons. No other study gives a verdict, so neither is the outlier.
3 papers rank ESM-2 above ESM3 in 20 of their 59 comparisons and ESM3 higher in 32; 7 are ties. Of those papers, ProteinGym ranks ESM3 higher on its own (18 of 19).