notesum.ai

Published at December 5

Does your model understand genes? A benchmark of gene properties for biological and text models

cs.AI

Released Date: December 5, 2024

Authors: Yoav Kan-Tor¹, Michael Morris Danziger¹, Eden Zohar¹, Matan Ninio¹, Yishai Shimoni¹

Aff.: ¹IBM Research - Israel

Arxiv: http://arxiv.org/pdf/2412.04075v1

Refer to caption

Model	Input type	Model type	Num of params	Output size
MTEB-L	Text	Transformer	7.1B	4,096
MTEB-S	Text	Transformer	109M	1024
MPNET	Text	Transformer	420 M	768
Bag-of-words	Text	Non-parametric	-	1,024
CellPLM	ScRNA-seq	Transformer	85M	1024
Geneformer	ScRNA-seq	Transformer	10.3M	256
ScGPT-H	ScRNA-seq	Transformer	51M	512
ScGPT-B	ScRNA-seq	Transformer	39M	512
Gene2Vec	Bulk RNA-seq	Word2Vec	5M	200
DNABERT-2	Base pair sequence	Transformer	117M	768
ESM-2	Protein sequence	Transformer	3B	2560