LT2 replaces quadratic attention in looped transformers with linear-time mechanisms, demonstrating synergistic gains from iteration and hybrid variants that achieve standard looped quality at linear cost or better performance with minimal full attention.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
LT2: Linear-Time Looped Transformers
LT2 replaces quadratic attention in looped transformers with linear-time mechanisms, demonstrating synergistic gains from iteration and hybrid variants that achieve standard looped quality at linear cost or better performance with minimal full attention.