A group attention mechanism that keeps important tokens and merges the rest cuts long-context decoding cost, but the theoretical case for grouping is only weakly supported.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Curse of High Dimensionality Issue in Transformer for Long-context Modeling
A group attention mechanism that keeps important tokens and merges the rest cuts long-context decoding cost, but the theoretical case for grouping is only weakly supported.