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.
Wu et al. 2025, Additional file 3 (Supplementary Tables S1-S13) ranks Geneformer above scGPT in 3 of 4 comparisons; other studies rank scGPT above Geneformer in 95 of 98.
3 papers rank scGPT above Geneformer in 95 of their 100 comparisons and Geneformer higher in 3; 2 are ties. Of those papers, Deep-learning-based gene perturbation… ranks Geneformer higher on its own (2 of 2). Wu et al. 2025 ranks Geneformer above scGPT in 3 of its 4 comparisons and scGPT higher in 1.
Deep-learning-based gene perturbation effect prediction does not yet outperform simple linear baselines ranks Geneformer above scGPT in 2 of 2 comparisons; other studies rank scGPT above Geneformer in 96 of 100.
3 papers rank scGPT above Geneformer in 96 of their 102 comparisons and Geneformer higher in 4; 2 are ties. Of those papers, Wu et al. 2025 ranks Geneformer higher on its own (3 of 4). Deep-learning-based gene perturbation… ranks Geneformer above scGPT in all 2 of its comparisons.