FOCUS, a Signum-based optimizer with an attraction force toward the EMA of parameters, is claimed to yield faster and more stable GPT-2 pretraining than Adam.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
FOCUS: First Order Concentrated Updating Scheme
FOCUS, a Signum-based optimizer with an attraction force toward the EMA of parameters, is claimed to yield faster and more stable GPT-2 pretraining than Adam.