specific method · filed under model architecture
IndexShare
Reuses one indexer across multiple sparse-attention layers.
- sources
- 3
- model
- 1
- lab adopt it
- 1
- strongest
- core
How sources treat it
One count per evidence span, weakest treatment to strongest.
Documented in
Further reading
Picked by hand, not extracted: where to read more, not evidence for anything on this page.
- IndexShare - LLM Architecture Gallery explainer sebastianraschka.comexplains IndexShare as used in GLM-5.2
Evidence
3 spans quoted from the sources, strongest treatment first.
the most technically significant innovation in GLM-5.2 is a mechanism Z.ai calls IndexShare
We propose IndexShare, which reuses the same indexer across every four sparse attention layers, reducing per-token FLOPs by 2.9× at a 1M context length.
We propose IndexShare, which reuses the same indexer across every four sparse attention layers, reducing per-token FLOPs by 2.9× at a 1M context length.
Filed alongside
Other methods under model architecture :: token mixer :: sparse attention :: sparse attention indexer.