en.wikipedia.org/w/index.php?title=Evo_(AI)&diff=prev&oldid=1369473152

Official sources describe Evo as being built on StripedHyena, not the Transformer architecture that gives GPTs their name.

• The Evo paper states: "StripedHyena is a deep signal processing architecture" and "We pretrained Evo ... with the StripedHyena architecture".
• Arc Institute's Evo 2 paper similarly says "Evo 2 uses StripedHyena 2, a convolutional multi-hybrid architecture" and explicitly compares it against Transformer baselines.

So labeling Evo as a "Generative pre-trained transformer" is architecturally incorrect: it is a genomic foundation model based on the StripedHyena family, not a GPT/Transformer.

Official sources consistently name the original Evo training dataset OpenGenome, not OpenGenome1.

• Arc Institute's original Evo announcement says the 300B-token dataset for Evo is "called OpenGenome".
• The Evo 2 materials describe the new dataset as OpenGenome2 and say it was created by expanding "the OpenGenome pretraining dataset, used to train Evo 1".

That makes the article's use of "OpenGenome1" inaccurate: the public name used by the project is OpenGenome.

The sentence misdescribes what was in Evo 1's training data.

Arc Institute's description of the original dataset says it was OpenGenome, a 300B-token collection of prokaryotic and phage genomes. That does not make phages "single cell organisms". Credible biomedical references define bacteriophages as viruses that infect bacteria.

So the article's wording is biologically incorrect: bacteria are single-celled organisms, but phages are viruses, not single-celled organisms.