Early-Signal Detector

method · first seen 13 Oct 2026

context parallelism

Also: prefill context parallelism

Distributing long sequences across devices by splitting the sequence dimension for attention computation.

Mentions per week

0219 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Earliest sources

Latest sources

Mentioned together with

FlashAttentionsparse attention maskslatent attentionKV cache offloadingspeculative decoding