Linear attention mechanisms and recurrent alternatives to softmax attention, including delta-rule and state-space models
| Name | Categories | Topics | References | Credibility | CreateDate | UpdateDate | Title | RelatedNotes | NavigationOrder | Eleventy | TemplateEngineOverride | Created |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| 20260629_carve_content_aware_recurrent_value_efficiency | paper_analysis | memory_mechanism, recurrent_neural_networks, llm, reasoning, linear_attention | carve, deltanet | 100 | CARVE: Content-Aware Recurrent with Value Efficiency for Chunk-Parallel Linear Attention | Gated DeltaNet-2: Decoupling Erase and Write in Linear Attention | [object Object] | [object Object] | md |