Early-Signal Detector

architecture · first seen 5 Oct 2026

linear attention

Also: linear transformers, linear recurrent neural network

Attention mechanism that replaces softmax with a kernel so cost scales linearly with sequence length.

Mentions per week

01319 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Rank 63 on 16 Nov 2026 · 24 mentions in 28 days · 24 independent groups · papers → code · first seen 6 weeks ago · acceleration -0.18

Earliest sources

Latest sources

Mentioned together with

Gated DeltaNetgated delta rulein-context learningsparse attentionrecurrent state quantizationKV cache quantizationhybrid attention-recurrent language modelsstate space modelsvideo world modelhybrid attention models