Model techniques map
Techniquesmodel architecturetoken mixersparse attentionsparse attention indexer

specific method · filed under model architecture

Hierarchical Sparse Indexer

Restricts later indexing stages to a candidate pool selected by an earlier indexer, reducing repeated scoring.

Also called Hierarchical sparse indexing.

sources
2
model
1
lab adopt it
1
strongest
core

How sources treat it

One count per evidence span, weakest treatment to strongest.

used 2core 3

Documented in

Evidence

5 spans quoted from the sources, strongest treatment first.

a Hierarchical Sparse Indexer restricts later indexers to a candidate pool selected by an earlier indexer, further reducing the number of entries scored per query.

coreinference servingin DeepSeek-V4.1-FlashDeepSeek

the first layer assigned to Full Mode constructs a candidate pool that later re-indexing layers use as their search domain.

coreinference servingin DeepSeek-V4.1-FlashDeepSeek

We therefore introduce the Hierarchical Sparse Indexer, which is used only in the decoder of CED to reduce this repeated scoring during decode.

coreunclearin DeepSeek-V4.1-FlashDeepSeek

For the Hierarchical Sparse Indexer, we select a maximum of 2,048 blocks with 8 positions, yielding up to 16,384 candidate positions in total.

usedunclearin DeepSeek-V4.1-FlashDeepSeek

In the decoder, a Hierarchical Sparse Indexer further restricts later indexing layers to a candidate pool constructed by the first Full Mode layer, bounding deeper indexer cost independently of context length.

usedmodel architecturein DeepSeek-V4.1-FlashDeepSeek

Filed alongside

Other methods under model architecture :: token mixer :: sparse attention :: sparse attention indexer.