Geo News
Community curated by people like you
LatestAICryptoHealthWorld AffairsUS Politics

Recurrent Neural Network stories

May 24, 2026

NVIDIA Releases Gated DeltaNet-2 Linear Attention Architecture

NVIDIA AI released Gated DeltaNet-2, a linear attention layer that decouples erase and write operations in the delta rule, addressing memory editing challenges in fixed-size recurrent states without scrambling existing associations.

May 24, 2026·1 source
00

Top claims

  • ▪Gated DeltaNet-2 has 1.3 billion parameters
  • ▪Linear attention squeezes the unbounded KV cache into a fixed-size recurrent state
  • ▪NVIDIA's Gated DeltaNet-2 decouples erase and write operations into a channel-wise erase gate b_t on the key axis and a channel-wise write gate w_t on the value axis

Subtopics

AI foundation models1AI research & benchmarks1Attention Mechanism1

Related timelines

Congress

108 stories

Iran War

209 stories

Payments

118 stories

Russia-Ukraine war

136 stories

Trump administration

116 stories