mlsanityData to 14 Oct 2026

CacheLink: Functional KV Transport for Zero-Prefill Cross-Model Context Handoff

arXiv 2610.15509 ↗First posted 11 Oct 2026cs.LG cs.AIcited by 0

On the map: Efficient LLM inference and serving → KV cache optimization

Abstract (arXiv)

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. Existing affine transport methods show that receiver KV states can be substantially reconstructed from sender caches. However, through extensive analyses, we find that reconstruction fidelity is a poor proxy for receiver utility. Similar K/V reconstruction errors can lead to substantially different attention patterns, predictive distributions, and downstream accuracy, because conventional reconstruction objectives overlook the query-dependent functional importance of K and V in receiver attention. We thus introduce CACHELINK, which learns a coupled K/V correction through the receiver behavior directly. Across four independently trained sender--receiver pairs and seven evaluation panels, CACHELINK improves the equal-weight mean over all methods for every pair. In matched diagnostics, CACHELINK changes K/V squared error by less than 0.4% while substantially changing receiver attention and next-token behavior. At 32K context, it reduces handoff latency by up to 7.27× relative to receiver re-prefill. 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.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in KV cache optimization, one of 12 directions in Efficient LLM inference and serving. Builds on 35 of its 38 references that are in the map. First posted on arXiv 11 Oct 2026.

Builds on

35 of 38 references.

Closest ancestors · 8

Same direction, newest first

  1. 2026CacheBridge: Efficient Cross-Model KV Cache Transfer1 Sep
  2. 2026Cross-Model KV Cache Transfer in LLM Families: A Closed-Form Linear Mapping for Prefill Reuse4 Aug
  3. 2025Cache-to-Cache: Direct Semantic Communication Between Large Language Models3 Oct
  4. 2025IAM: Efficient Inference through Attention Mapping between Different-scale LLMs16 Jul
  5. 2024Quest: Query-Aware Sparsity for Efficient Long-Context LLM Inference16 Jun
  6. 2024CacheBlend: Fast Large Language Model Serving for RAG with Cached Knowledge Fusion26 May
  7. 2024SnapKV: LLM Knows What You are Looking for Before Generation22 Apr
  8. 2023Efficient Memory Management for Large Language Model Serving with PagedAttention12 Sep

Other references · 27

  1. 2025Qwen3 Technical Reportcited by 8,820
  2. 2018Can a Suit of Armor Conduct Electricity? A New Dataset for Open Book Question Answeringcited by 2,164
  3. 2024MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmarkcited by 1,848
  4. 2019Similarity of Neural Network Representations Revisitedcited by 1,755
  5. 2023SGLang: Efficient Execution of Structured Language Model Programscited by 1,141
  6. 2023FrugalGPT: How to Use Large Language Models While Reducing Cost and Improving Performancecited by 853
  7. 2017Crowdsourcing Multiple Choice Science Questionscited by 765
  8. 2024Yi: Open Foundation Models by 01.AIcited by 719
  9. 2024RouteLLM: Learning to Route LLMs with Preference Datacited by 682
  10. 2024KIVI: A Tuning-Free Asymmetric 2bit Quantization for KV Cachecited by 637
  11. 2024LongBench v2: Towards Deeper Understanding and Reasoning on Realistic Long-context Multitaskscited by 379
  12. 2023Prompt Cache: Modular Attention Reuse for Low-Latency Inferencecited by 268
15 more
  1. 2024RouterBench: A Benchmark for Multi-LLM Routing Systemcited by 238
  2. 2024BABILong: Testing the Limits of LLMs with Long Context Reasoning-in-a-Haystackcited by 215
  3. 2021Revisiting Model Stitching to Compare Neural Representationscited by 209
  4. 2023AutoMix: Automatically Mixing Language Modelscited by 125
  5. 2026Ministral 3cited by 114
  6. 2022Language Model Cascadescited by 107
  7. 2024LongAlign: A Recipe for Long Context Alignment of Large Language Modelscited by 106
  8. 2024Preble: Efficient Distributed Prompt Scheduling for LLM Servingcited by 70
  9. 2025Oolong: Evaluating Long Context Reasoning and Aggregation Capabilitiescited by 40
  10. 2025Transferring Linear Features Across Language Models With Model Stitchingcited by 21
  11. 2026Latent Space Communication via K-V Cache Alignmentcited by 17
  12. 2024MMLU-Pro+: Evaluating Higher-Order Reasoning and Shortcut Learning in LLMscited by 16
  13. 2026EvoRoute: Experience-Driven Self-Routing LLM Agent Systemscited by 16
  14. 2026ICaRus: Identical Cache Reuse for Efficient Multi Model Inferencecited by 8
  15. 2026PrefillShare: A Shared Prefill Module for KV Reuse in Multi-LLM Disaggregated Servingcited by 4

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2026To Cache or Not to Cache: Understanding and Mitigating Safety Incompatibility in Cross-Model KV Cache Reuse
  2. 2026A Universal Context-Reuse Layer for Cross-Model KV Sharing
  3. 2026ImpactHO: Importance-Aware KV Cache Transfer for Multi-User Edge LLM Handover
  4. 2026Prefill-Free Cross-Family KV Cache Transfer for Heterogeneous Multi-Agent LLMs
  5. 2026EchoPress: Query-Agnostic KV Cache Pruning via Virtual Context Reconstruction
  6. 2026Receiver-Conditioned Latent Communication gives 94% CacheBack
  7. 2026CORE: COverage CAlibration and Evicted-Mass REdistribution for KV Cache
  8. 2025KV Cache Recycling to Expand Usable Context Capacity in Low Parameter LLMs