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.
Zero-shot benchmarking of RNA language models in structural, functional, and evolutionary learning ranks RNA-FM above RNAErnie in 28 of 36 comparisons; 2 of 2 other studies rank RNAErnie above RNA-FM.
2 papers rank RNAErnie above RNA-FM in all 6 comparisons that are not ties (13 are ties). Zero-shot benchmarking of RNA language… ranks RNA-FM above RNAErnie in 28 of its 36 comparisons and RNAErnie higher in 8.
mRNABench: A curated benchmark for mature mRNA property and function prediction ranks DNABERT-2 above RNA-FM in 7 of 7 comparisons; 2 of 2 other studies rank RNA-FM above DNABERT-2.
2 papers rank RNA-FM above DNABERT-2 in all 30 comparisons that are not ties (4 are ties). MRNABench ranks DNABERT-2 above RNA-FM in all 7 comparisons that are not ties (3 are ties).
Benchmarking pre-trained genomic language models for RNA sequence-related predictive applications ranks SpliceBERT above RNA-FM in 3 of 3 comparisons; Zero-shot benchmarking of RNA language models in structural, functional, and evolutionary learning ranks RNA-FM above SpliceBERT in 31 of 36 comparisons. No other study gives a verdict, so neither is the outlier.
2 papers rank RNA-FM above SpliceBERT in 31 of their 51 comparisons and SpliceBERT higher in 8; 12 are ties. Of those papers, Benchmarking pre-trained genomic… ranks SpliceBERT higher on its own (3 of 15).