Metraipolis › archive › dc_ecbed1ca71fd

KVCMAS: Efficient KV cache Correction for Shared Context in Multi-Agent Systems — arXiv:2609.34060

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.34060
License (as archived): CC0-1.0
Archived: 2026-10-07T22:37:58.674Z
SHA-256 of the archived text: f5cd2629fb50ace633955d426a6d8242003d9a21e8861cfb9559d9491d4ed091
Stable URL of this page: https://metraipolis.com/documento/dc_ecbed1ca71fd

How to cite

KVCMAS: Efficient KV cache Correction for Shared Context in Multi-Agent Systems — arXiv:2609.34060 — arXiv (en), https://arxiv.org/abs/2609.34060 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ecbed1ca71fd (Metraipolis archive dc_ecbed1ca71fd, sha256 f5cd2629fb50ace633955d426a6d8242003d9a21e8861cfb9559d9491d4ed091)

BibTeX

@misc{dcecbed1ca71fd,
  title = {KVCMAS: Efficient KV cache Correction for Shared Context in Multi-Agent Systems — arXiv:2609.34060},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.34060},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ecbed1ca71fd; sha256 f5cd2629fb50ace633955d426a6d8242003d9a21e8861cfb9559d9491d4ed091},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34060. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Prompt-specialized multi-agent systems enable multiple agents to share a model while performing complementary roles to solve complex tasks. However, agent-specific prefixes change the KV cache generated for the same shared context, causing each agent to repeatedly prefill the growing context and construct a separate cache with high computation and memory overhead. Selective recomputation reduces this redundancy but still retains substantial model execution, while existing delta correction methods either support only recurring context relations or maintain memory-intensive online correction states for dynamically changing context. For first seen shared context, these methods also construct a reference cache outside the agent workflow, and an approximate correction at the first agent affects the outputs passed to subsequent agents. We present KVCMAS, an online KV cache correction framework that represents cross-agent cache deviations using compact low-rank states and seamlessly chains corrections along the agent workflow without an additional reference prefill. This design supports dynamically changing shared context while preserving an exact first-agent cache. Across multiple language and vision-language workloads, KVCMAS matches or improves the accuracy of prior KV cache sharing methods while achieving the lowest TTFT under highly concurrent serving. Under controlled serving traces, it provides a 2.0x TTFT speedup over inference without KV cache sharing and reduces peak GPU memory by up to 3.7x relative to a prior KV cache correction method. These results establish KVCMAS as an accurate and scalable KV cache sharing approach for prompt-specialized multi-agent serving.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.34060 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T22:14:28.103Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).