Omer Levy
Omer Levy is a researcher whose papers cover word embeddings, pre-training methods and language model benchmarks, all of which concern how AI systems handle written language. In 2014 he and Yoav Goldberg published a NIPS paper on implicit matrix factorization in word embeddings, and in 2016 he completed a PhD at Bar-Ilan University. In 2020 he joined Tel Aviv University and co-authored SpanBERT and BART, and in 2021 he co-authored a study of transformer feed-forward layers as key-value memories. SCROLLS, a standardized comparison over long language sequences, came out in 2022 with him among its authors. As of October 2026, he is a Scholar in the ELLIS Tel Aviv unit at Tel Aviv University.
- Now at
- Tel Aviv University
- Works
- 142
- Citations
- 49,574
- h-index
- 48
Publication numbers from OpenAlex, as of 15 October 2026.
Timeline
-
2014
Word embeddings as matrix factorization
In 2014 he and Yoav Goldberg publish a paper titled Neural Word Embedding as Implicit Matrix Factorization at NIPS.
-
2016
A PhD on word embeddings
He studies at the Technion and completes a PhD at Bar-Ilan University in 2016 with Yoav Goldberg as an advisor.
-
2020
Joining Tel Aviv University
Tel Aviv University becomes his employer in 2020.
-
2020
SpanBERT and BART
SpanBERT and BART, two pre-training methods, list him as a co-author in 2020.
-
2021
Feed-forward layers as memories
The 2021 conference paper Transformer Feed-Forward Layers Are Key-Value Memories lists him as a co-author.
-
2022
SCROLLS for long sequences
SCROLLS, a standardized comparison over long language sequences, comes out in 2022 with him among its authors.
-
2023
LMentry, a language model benchmark
The 2023 LMentry paper, written with Avia Efrat and Or Honovich, presents a language model benchmark of elementary tasks.
-
As of October 2026
An ELLIS Scholar in Tel Aviv
He is a Scholar in the ELLIS Tel Aviv unit at Tel Aviv University.
Sources and credits
Photo: We have no published photo of Omer Levy yet. Send us a free-licensed photo.
- Neural Word Embedding as Implicit Matrix Factorization
- Omer Levy - The Mathematics Genealogy Project
- Omer Levy - Wikidata
- SpanBERT: Improving Pre-training by Representing and Predicting Spans - ACL Anthology
- BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension - ACL Anthology
- Transformer Feed-Forward Layers Are Key-Value Memories - ACL Anthology
- SCROLLS: Standardized CompaRison Over Long Language Sequences - ACL Anthology
- LMentry: A Language Model Benchmark of Elementary Language Tasks - ACL Anthology
- Omer Levy | European Laboratory for Learning and Intelligent Systems