{"slug": "cross-model-kv-cache-transfer-in-llm-families-a-closed-form-linear-mapping-for", "title": "Cross-model KV cache transfer in LLM families: a closed-form linear mapping for prefill reuse", "summary": "Researchers proposed cross-model KV cache transfer, a method that reuses a source model's key-value cache to skip prefill when swapping between different-sized models in a family, and found a closed-form linear mapping retains 73-98% of standalone-prefill accuracy on four of six pairs tested across three families, with the mapper running 2.7-25x faster than re-prefill. On Qwen3 14B to 32B, one source layer explains 56% of variance in target keys and 32% in values, rising to 79% and 65% with multiple layers, and a nonlinear MLP recovers up to +37 percentage points HellaSwag retention on the two failing pairs.", "body_md": "# Computer Science > Machine Learning\n\n  [Submitted on 4 Aug 2026]\n\n# Title:Cross-Model KV Cache Transfer in LLM Families: A Closed-Form Linear Mapping for Prefill Reuse\n\n[View PDF](/pdf/2608.03893)\n\n[HTML (experimental)](https://arxiv.org/html/2608.03893v1)\n\nAbstract:Production deployments often swap between different-sized models in a family for cost-quality cascading, mid-conversation switching, and routing, and each swap forces the receiver to repay the prefill from scratch. We propose cross-model KV cache transfer, where the receiver reuses the source's KV cache, skipping prefill. We find that cross-model KV has substantial linear structure across matched-KV pairs, where source and target share KV head count and per-head dimension. On Qwen3 14B->32B, one source layer explains 56% of variance in the target's keys and 32% in values, rising to 79% and 65% with multiple source layers. Building on this, we design a closed-form ridge mapper that operates per head and proceeds in three steps. First, for each target layer we select the top-k most predictive source layers and concatenate their KV as input. Second, we strip RoPE from the keys before mapping, so the fit is position-free and reusable across context lengths. Third, we fit ridge regression on a small calibration set of 500 FineWeb-Edu sequences of 1,024 tokens each. Surprisingly, across six pairs in three families, this linear mapper retains 73-98% of the receiver's standalone-prefill accuracy on four pairs, while two degrade sharply. A nonlinear MLP recovers up to +37 pp HellaSwag retention on the failures. The mapper runs 2.7-25x faster than re-prefill and remains stable across multi-turn handoff, making cross-model KV cache transfer practical.\n    \n\n### References & Citations\n\nLoading...\n\n# Bibliographic and Citation Tools\n\nBibliographic Explorer \n\n*(*[What is the Explorer?](https://info.arxiv.org/labs/showcase.html#arxiv-bibliographic-explorer))\nConnected Papers \n\n*(*[What is Connected Papers?](https://www.connectedpapers.com/about))\nLitmaps \n\n*(*[What is Litmaps?](https://www.litmaps.co/))\nscite Smart Citations \n\n*(*[What are Smart Citations?](https://www.scite.ai/))\n# Code, Data and Media Associated with this Article\n\nalphaXiv \n\n*(*[What is alphaXiv?](https://alphaxiv.org/))\nCatalyzeX Code Finder for Papers \n\n*(*[What is CatalyzeX?](https://www.catalyzex.com))\nDagsHub \n\n*(*[What is DagsHub?](https://dagshub.com/))\nGotit.pub \n\n*(*[What is GotitPub?](http://gotit.pub/faq))\nHugging Face \n\n*(*[What is Huggingface?](https://huggingface.co/huggingface))\nScienceCast \n\n*(*[What is ScienceCast?](https://sciencecast.org/welcome))\n# Demos\n\n# Recommenders and Search Tools\n\nInfluence Flower \n\n*(*[What are Influence Flowers?](https://influencemap.cmlab.dev/))\nCORE Recommender \n\n*(*[What is CORE?](https://core.ac.uk/services/recommender))\nIArxiv Recommender\n\n*(*[What is IArxiv?](https://iarxiv.org/about))\n# arXivLabs: experimental projects with community collaborators\n\narXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.\n\nBoth individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.\n\nHave an idea for a project that will add value for arXiv's community? [**Learn more about arXivLabs**](https://info.arxiv.org/labs/index.html).", "url": "https://wpnews.pro/news/cross-model-kv-cache-transfer-in-llm-families-a-closed-form-linear-mapping-for", "canonical_source": "https://arxiv.org/abs/2608.03893", "published_at": "2026-09-09 09:00:00+00:00", "updated_at": "2026-09-09 11:13:37.786269+00:00", "lang": "en", "topics": ["machine-learning", "large-language-models", "ai-research"], "entities": ["Qwen3", "FineWeb-Edu", "HellaSwag", "arXiv"], "alternates": {"html": "https://wpnews.pro/news/cross-model-kv-cache-transfer-in-llm-families-a-closed-form-linear-mapping-for", "markdown": "https://wpnews.pro/news/cross-model-kv-cache-transfer-in-llm-families-a-closed-form-linear-mapping-for.md", "text": "https://wpnews.pro/news/cross-model-kv-cache-transfer-in-llm-families-a-closed-form-linear-mapping-for.txt", "jsonld": "https://wpnews.pro/news/cross-model-kv-cache-transfer-in-llm-families-a-closed-form-linear-mapping-for.jsonld"}}