implementation detail · filed under model architecture
Kimi Delta Attention with input-dependent full-rank output gate
A Kimi Delta Attention implementation that replaces a low-rank output-gate parameterization with an input-dependent full-rank projection.
Also called Input-dependent full-rank output gating, input-dependent full-rank projection.
- source
- 1
- model
- 1
- lab adopt it
- 1
- strongest
- used
How sources treat it
One count per evidence span, weakest treatment to strongest.
used 1
Documented in
Evidence
1 span quoted from the sources, strongest treatment first.
Kimi K3 changes KDA’s output gate from the low-rank parameterization used by Kimi Linear [63] to an input-dependent full-rank projection.
usedmodel architecturein Kimi K3Moonshot AI
Filed alongside
Other methods under model architecture :: token mixer :: linear attention & state space :: gated delta network.
Gated DeltaNetKimi Delta AttentionGated DeltaNet–sparse MoE hybridKDAGated DeltaNet with bounded sigmoid output gateGated DeltaNet with reduced KV-head configurationHybrid linear attentionKimi Delta Attention and Attention Residuals architectureKimi Delta Attention with lower-bounded log-decaySimpleGDN