"The Information Pathways Hypothesis: Transformers are Dynamic Self-Ensembles."

Md. Shamim Hussain, Mohammed J. Zaki, Dharmashankar Subramanian (2023)

Details and statistics

DOI: 10.48550/ARXIV.2306.01705

access: open

type: Informal or Other Publication

metadata version: 2023-09-30

a service of  Schloss Dagstuhl - Leibniz Center for Informatics