Ofir Press
BuilderPrinceton
@ofirpress
Researcher at Princeton. Inventor of ALiBi positional encoding enabling length generalization in transformers without position embeddings.
Skills
Looking For
Research Foundations (2)
Projects (1)
ALiBi Positional Encoding
Attention with Linear Biases enables transformers to generalize to longer sequences than seen during training without learned position embeddings.
Built on research:
AI Suggested Researchers
Researchers whose work may be relevant to your projects (auto-detected)
Hannaneh Hajishirzi
University of Washington / Allen AI
Kyunghyun Cho
New York University / Genentech
Aaron Courville
Mila / Université de Montréal
Luke Zettlemoyer
University of Washington / Meta AI
AI Suggested Papers
Papers that may have inspired your projects (auto-detected by domain & keyword analysis)
For ALiBi Positional Encoding:
BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding
NAACL 2019 · 2019 · 108,124 citations
COMET: Commonsense Transformers for Automatic Knowledge Graph Construction
ACL 2019 · 2019 · 1,500 citations
Language Models are Few-Shot Learners
NeurIPS 2020 · 2020 · 18,500 citations