Prefill-Free Cross-Family KV Cache Transfer for Heterogeneous Multi-Agent LLMs

类型：arxiv
标识：2609.32259
链接：https://arxiv.org/abs/2609.32259
主分类：agent
形态：method
TLDR：Recent multi-agent LLM systems increasingly combine heterogeneous models for specialized agent roles. However, text-based communication requires each receiver to prefill shared context already processed by the sender. Reusing the sender's key-value (KV) cache avoids this redundancy, but prefill-free transfer across model families must handle differences in tokenization, model depth, and KV representations. To address these issues, we propose HeteroFold, a prefill-free cross-family KV cache transfer method that keeps both the sender and receiver frozen. HeteroFold aligns model structures, maps
副分类：llm-infra
待LLM分类：否
来源文件：
/inbox/tom/_candidates/2026-10-02-agent-rag-longcontext-candidates.json

相关 · 按语义匹配

笔记 flyP 轻量精读 · 2026-10-01 22:50

2026-10-01

笔记 Tom 文献雷达 · Agent + RAG + Long-Context · 2026-08-13

2026-08-13

笔记知识库草稿：KV Cache 系统工程 · Inference Stack 商业化 · ChromaDB 安全警报

2026-06-11

笔记 Substack · α-signal · RAG and Long Context Aren't Enough for Agent Memory. δ-mem Is a Third Option

2026-09-09

笔记 flyP · 22:50 短审稿 · KV Cache Reuse 评估方法学(`arXiv:2609.31415`)

2026-09-30

笔记知识库草稿 · Jay · 2026-08-01 19:50（傍晚高频）

2026-08-01

arXiv

2609.32259

TLDR

Recent multi-agent LLM systems increasingly combine heterogeneous models for specialized agent roles. However, text-based communication requires each receiver to prefill shared context already processed by the sender. Reusing the sender's key-value (KV) cache avoids this redundancy, but prefill-free transfer across model families must handle differences in tokenization, model depth, and KV representations. To address these issues, we propose HeteroFold, a prefill-free cross-family KV cache transfer method that keeps both the sender and receiver frozen. HeteroFold aligns model structures, maps