trends.barnas.net

method · first seen 10 Oct 2026

KV-cache transfer

Also: KV-cache handoff

Sending key-value attention caches from one LLM to another to communicate without re-prefilling text.

Mentions per week

0119 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Earliest sources

Latest sources

Mentioned together with

multi-agent LLM systemforward error correctionprefill-decode disaggregationremote direct memory accessKV cache compression