Home / AI Research / Accelerating Mixture of Block Attention via Dynamic Routing, KV-Centric Execution and Runtime Pipelining
ACAI Research
Accelerating Mixture of Block Attention via Dynamic Routing, KV-Centric Execution and Runtime Pipelining
by tencent · Yingqi Shi, Junru Lu, Yun Liu, Di Yin, Xing Sun
Mixture of Block Attention (MoBA) reduces long-context computation by selecting key-value (KV) blocks for each query. FlashMoBA implements MoBA efficiently but retains a fixed Top-k budget and query-centric schedule. Fixed Top-k applies one budget across different score distributions and couples routing to task granularity; query-range ownership fragments K/V reuse when several ranges select the same column. We present LightningMoBA, which co-designs selection and execution through dynamic routing, KV-centric ownership, and runtime pipelining. Dynamic routing thresholds MoBA's per-query block scores and directly emits variable-cardinality sparse metadata. KV-centric ownership groups…
Upvotes18
Comments3
Published18 Nov 2026