Genomic sequence modeling on BEND
0.67Gene Finding MCCWisteria
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| WisteriaHidden dimension=768, Pre-training sequence length=131,072 bp2026.05 | 0.67 | 5 | 85 | 84 | 93 | 62 | 75 | |
| NT-v22026.05 | 0.64 | 5 | 80 | 76 | 91 | 48 | 48 | |
| GENA-LMArchitecture=BigBird configuration, Layers=12, Hidden dimension=7682026.05 | 0.52 | 3 | 76 | 78 | 91 | 49 | 55 | |
| DNABERT-22026.05 | 0.43 | 3 | 81 | 78 | 90 | 49 | 51 | |
| HyenaDNAModel size=large model configuration, Supported sequence length=up to 1,000,000 bp, Tokenization=single-nucleotide resolution2026.05 | 0.35 | 3 | 84 | 76 | 91 | 51 | 45 | |
| DNABERTArchitecture=BERT base, Layers=12, Hidden dimension=768, Tokenization=6-mer, Input window size=512 bp2026.05 | 0.2 | 3 | 85 | 79 | 91 | 60 | 56 |