implementation detail · filed under model architecture
Kimi Delta Attention with lower-bounded log-decay
A Kimi Delta Attention decay parameterization that uses a scaled sigmoid to bound log-decay from below.
Also called lower-bounded decay.
- 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 instead uses a scaled sigmoid to bound the log-decay from below
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 input-dependent full-rank output gateSimpleGDN