A two-layer transformer with relative positional encoding keeps its induction head active across the whole sequence, while absolute positional encoding loses it in the second half.
Learning patterns and pattern se- quences by self-organizing nets of threshold el- ements.IEEE Transactions on computers, 100 (11):1197–1206, 1972
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CL 1years
2024 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Rethinking Associative Memory Mechanism in Induction Head
A two-layer transformer with relative positional encoding keeps its induction head active across the whole sequence, while absolute positional encoding loses it in the second half.