GENEB benchmark reveals unstable model rankings across task categories, modest inconsistent gains from scale, and greater importance of architecture and pretraining alignment over parameter count.
Understanding the natural language of DNA using encoder–decoder foundation models with byte-level precision , volume=
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
GENEB: Why Genomic Models Are Hard to Compare
GENEB benchmark reveals unstable model rankings across task categories, modest inconsistent gains from scale, and greater importance of architecture and pretraining alignment over parameter count.