CacheLink: Functional KV Transport for Zero-Prefill Cross-Model Context Handoff
KV caches make long-context serving efficient by reusing past computation, but model switching breaks this reuse and forces the receiver to prefill the context again.
ProofPaper ↗
Key points
- We study zero-prefill cross-model context handoff: translating a sender cache into a receiver-compatible state so that the receiver can continue directly without prefilling same context.
- Existing affine transport methods show that receiver KV states can be substantially reconstructed from sender caches.
- We thus introduce CACHELINK, which learns a coupled K/V correction through the receiver behavior directly.
- Our findings verify that reconstruction fidelity alone is insufficient and that preserving the receiver computation induced by the transported cache is important for effective cross-model KV transport.
Sources (1)
- [1]CacheLink: Functional KV Transport for Zero-Prefill Cross-Model Context HandoffarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 11, 10:19 AM
KV caches make long-context serving efficient by reusing past computation, but model switching breaks this reuse and forces the receiver to prefill the context again.
We study zero-prefill cross-model context handoff: translating a sender cache into a receiver-compatible state so that the receiver can continue directly without prefilling same context.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 11, 2026OrcaRouter Releases OrcaCyber Zero 1.5 Cybersecurity Model With 1M Context
- Oct 11, 2026Apertus 1.5 Release Notes
- Oct 11, 2026History Replays Itself: Unleashing Long-Context Capability in Linear Attention
- Oct 10, 2026Humans vs. Machines: A Comparison With AI Agents in Software Reverse Engineering
- Oct 9, 2026Microsoft's Decision-1 model enters the fast-growing AI decision model race
- Oct 8, 2026[AINews] not much happened today