TLinFormer compresses long history into a fixed-size context state to make each full forward pass linear in sequence length, but the attention is not exact and per-token generation still costs O(N).
Abo-Zahhad, Aziza I
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
Rethinking Transformer Connectivity: TLinFormer, A Path to Exact, Full Context-Aware Linear Attention
TLinFormer compresses long history into a fixed-size context state to make each full forward pass linear in sequence length, but the attention is not exact and per-token generation still costs O(N).