2026
CLM-AI-003624= Theta(L d_h)ASYMPTOTIC BOUNDThe cited theorem establishes Theta(L d_h) under the recorded model.Source: Fast Transformer Decoding: One Write-Head is All You Need — Multi-query-attention KV-cache scaling ↗
PROVENA single shared key/value head makes per-layer KV cache linear in context length and head width, independent of query-head count.
The question, scope, and sources behind this Registry record.
Under the recorded transformer systems and scaling model, establish Theta(L d_h).
Current frontiers derived from accepted Claims.
A parameterized or asymptotic result under the current specification.
Assertions tied to evidence, attribution, and review.
The frontier as it changed over time.
Only accepted Claims matching the current specification contribute to the displayed bounds. Strict inequalities remain open; contradictory Claims require editorial review.
No accepted machine-checked reproductions are recorded for this Limit.
Limits Registry. LR-003624. Multi-query-attention KV-cache scaling. 2026.