# Knowledge Packs: Zero-Token Knowledge Delivery via KV Cache Injection

arXiv: https://arxiv.org/abs/2604.03270 · HTML: https://mlsanity.com/paper/2604.03270/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Placed in KV cache optimization, one of 12 directions in Efficient LLM inference and serving.
- Builds on 18 of its 21 references that are in the map.
- First posted on arXiv 22 Mar 2026.

## Placed in

- Efficient LLM inference and serving / [KV cache optimization](https://mlsanity.com/inference/kv-cache-optimization/)

## Builds on (most cited first, 18 of 18)

- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks](https://mlsanity.com/paper/2005.11401/) (2020)
- [Efficient Memory Management for Large Language Model Serving with PagedAttention](https://mlsanity.com/paper/2309.06180/) (2023)
- [RoFormer: Enhanced Transformer with Rotary Position Embedding](https://mlsanity.com/paper/2104.09864/) (2021)
- [Prefix-Tuning: Optimizing Continuous Prompts for Generation](https://mlsanity.com/paper/2101.00190/) (2021)
- [The Power of Scale for Parameter-Efficient Prompt Tuning](https://mlsanity.com/paper/2104.08691/) (2021)
- [HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering](https://mlsanity.com/paper/1809.09600/) (2018)
- [Locating and Editing Factual Associations in GPT](https://mlsanity.com/paper/2202.05262/) (2022)
- [REALM: Retrieval-Augmented Language Model Pre-Training](https://mlsanity.com/paper/2002.08909/) (2020)
- [Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection](https://mlsanity.com/paper/2310.11511/) (2023)
- [Inference-Time Intervention: Eliciting Truthful Answers from a Language Model](https://mlsanity.com/paper/2306.03341/) (2023)
- [Measuring and Narrowing the Compositionality Gap in Language Models](https://mlsanity.com/paper/2210.03350/) (2022)
- [Mass-Editing Memory in a Transformer](https://mlsanity.com/paper/2210.07229/) (2022)
- [Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questions](https://mlsanity.com/paper/2212.10509/) (2022)
- [SnapKV: LLM Knows What You are Looking for Before Generation](https://mlsanity.com/paper/2404.14469/) (2024)
- [H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models](https://mlsanity.com/paper/2306.14048/) (2023)
- [Scissorhands: Exploiting the Persistence of Importance Hypothesis for LLM KV Cache Compression at Test Time](https://mlsanity.com/paper/2305.17118/) (2023)
- [Agent Memory Below the Prompt: Persistent Q4 KV Cache for Multi-Agent LLM Inference on Edge Devices](https://mlsanity.com/paper/2603.04428/) (2026)

## Built on by (most cited first, 2 of 2)

- [DepthWeave-KV: Token-Adaptive Cross-Layer Residual Factorization for Long-Context KV Cache Compression](https://mlsanity.com/paper/2607.06523/) (2026)
- [FreqDepthKV: Frequency-Guided Depth Sharing for Robust KV Cache Compression in Long-Context LLM Inference](https://mlsanity.com/paper/2607.06519/) (2026)

## Abstract (arXiv)

RAG wastes tokens. We propose Knowledge Packs: pre-computed KV caches that deliver the same knowledge at zero token cost. For causal transformers, the KV cache from a forward pass on text F is identical to what a joint pass on F+q would produce - this follows directly from the causal mask. The equivalence is exact but fragile: wrong chat template formatting causes 6-7pp degradation, which we believe explains prior claims of KV outperforming RAG. With correct formatting: zero divergences across 700 questions on Qwen3-8B and Llama-3.1-8B, up to 95% token savings. The KV interface also enables behavioral steering that RAG cannot do. Because RoPE rotates keys but leaves values untouched, contrastive deltas on cached values can nudge model behavior while key arithmetic destroys coherence. The effect sits in mid-layer values (33-66%), independent directions are nearly orthogonal (cos~0) and compose, and both channels - knowledge and steering - run simultaneously at alpha<=0.7 without interference. No training, no weight modification.