# Cache Me If You Can: How Many KVs Do You Need for Effective Long-Context LMs?

arXiv: https://arxiv.org/abs/2506.17121 · HTML: https://mlsanity.com/paper/2506.17121/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- Builds on 56 of its 64 references that are in the map.
- First posted on arXiv 20 Jun 2025.

## Builds on (most cited first, 30 of 56)

- [Attention Is All You Need](https://mlsanity.com/paper/1706.03762/) (2017)
- [The Llama 3 Herd of Models](https://mlsanity.com/paper/2407.21783/) (2024)
- [DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning](https://mlsanity.com/paper/2501.12948/) (2025)
- [Mamba: Linear-Time Sequence Modeling with Selective State Spaces](https://mlsanity.com/paper/2312.00752/) (2023)
- [HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering](https://mlsanity.com/paper/1809.09600/) (2018)
- [Lost in the Middle: How Language Models Use Long Contexts](https://mlsanity.com/paper/2307.03172/) (2023)
- [TriviaQA: A Large Scale Distantly Supervised Challenge Dataset for Reading Comprehension](https://mlsanity.com/paper/1705.03551/) (2017)
- [Categorical Reparameterization with Gumbel-Softmax](https://mlsanity.com/paper/1611.01144/) (2016)
- [Efficiently Modeling Long Sequences with Structured State Spaces](https://mlsanity.com/paper/2111.00396/) (2021)
- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [Transformers are RNNs: Fast Autoregressive Transformers with Linear Attention](https://mlsanity.com/paper/2006.16236/) (2020)
- [GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers](https://mlsanity.com/paper/2210.17323/) (2022)
- [Fast Inference from Transformers via Speculative Decoding](https://mlsanity.com/paper/2211.17192/) (2022)
- [MS MARCO: A Human Generated MAchine Reading COmprehension Dataset](https://mlsanity.com/paper/1611.09268/) (2016)
- [SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Models](https://mlsanity.com/paper/2211.10438/) (2022)
- [GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints](https://mlsanity.com/paper/2305.13245/) (2023)
- [RULER: What's the Real Context Size of Your Long-Context Language Models?](https://mlsanity.com/paper/2404.06654/) (2024)
- [The Concrete Distribution: A Continuous Relaxation of Discrete Random Variables](https://mlsanity.com/paper/1611.00712/) (2016)
- [SGLang: Efficient Execution of Structured Language Model Programs](https://mlsanity.com/paper/2312.07104/) (2023)
- [DeepSeek-V2: A Strong, Economical, and Efficient Mixture-of-Experts Language Model](https://mlsanity.com/paper/2405.04434/) (2024)
- [When Not to Trust Language Models: Investigating Effectiveness of Parametric and Non-Parametric Memories](https://mlsanity.com/paper/2212.10511/) (2022)
- [SnapKV: LLM Knows What You are Looking for Before Generation](https://mlsanity.com/paper/2404.14469/) (2024)
- [Enhancing Chat Language Models by Scaling High-quality Instructional Conversations](https://mlsanity.com/paper/2305.14233/) (2023)
- [H2O: Heavy-Hitter Oracle for Efficient Generative Inference of Large Language Models](https://mlsanity.com/paper/2306.14048/) (2023)
- [RWKV: Reinventing RNNs for the Transformer Era](https://mlsanity.com/paper/2305.13048/) (2023)
- [Retentive Network: A Successor to Transformer for Large Language Models](https://mlsanity.com/paper/2307.08621/) (2023)
- [The NarrativeQA Reading Comprehension Challenge](https://mlsanity.com/paper/1712.07040/) (2017)
- [Learning Sparse Neural Networks through L0 Regularization](https://mlsanity.com/paper/1712.01312/) (2017)
- [Quest: Query-Aware Sparsity for Efficient Long-Context LLM Inference](https://mlsanity.com/paper/2406.10774/) (2024)
- [Model Tells You What to Discard: Adaptive KV Cache Compression for LLMs](https://mlsanity.com/paper/2310.01801/) (2023)

## Built on by (most cited first, 10 of 10)

- [Which Heads Matter for Reasoning? RL-Guided KV Cache Compression](https://mlsanity.com/paper/2510.08525/) (2025)
- [Elastic Attention: Test-time Adaptive Sparsity Ratios for Efficient Transformers](https://mlsanity.com/paper/2601.17367/) (2026)
- [LycheeDecode: Accelerating Long-Context LLM Inference via Hybrid-Head Sparse Decoding](https://mlsanity.com/paper/2602.04541/) (2026)
- [NOSA: Native and Offloadable Sparse Attention](https://mlsanity.com/paper/2510.13602/) (2025)
- [KV Admission: Learning What to Write for Efficient Long-Context LLM Inference](https://mlsanity.com/paper/2512.17452/) (2025)
- [LKV: End-to-End Learning of Head-wise Budgets and Token Selection for LLM KV Cache Eviction](https://mlsanity.com/paper/2605.06676/) (2026)
- [MachineLearningLM: Scaling Many-shot In-context Learning via Continued Pretraining](https://mlsanity.com/paper/2509.06806/) (2025)
- [FlashAccel: Leveraging High-Bandwidth Flash (HBF) for High-Throughput LLM Inference](https://mlsanity.com/paper/2607.10186/) (2026)
- [Prefix Sliding for efficient test-time scaling](https://mlsanity.com/paper/2608.26070/) (2026)
- [Flux Attention: Context-Aware Hybrid Attention for Efficient LLMs Inference](https://mlsanity.com/paper/2604.07394/) (2026)

## Abstract (arXiv)

Language models handle increasingly long contexts for tasks such as book summarization, but this leads to growing memory costs for the key-value (KV) cache. Many prior works have proposed ways of discarding KVs from memory, but their approaches are tailored to favorable settings, obscuring caveats like high peak memory and performance degradation, and a fair comparison between methods is difficult. In this paper, we propose the *KV footprint* as a unified metric, which accounts for both the amount of KV entries stored and their lifespan in memory. We evaluate methods based on the smallest footprint they attain while preserving performance in both long-context understanding and generation, with context lengths of up to 128K tokens. This metric reveals the high peak memory of prior KV eviction methods. One class of methods -- *post-fill eviction* -- has a high footprint due to being incompatible with eviction during pre-filling. We adapt these methods to be able to evict KVs during pre-filling, achieving substantially lower KV footprints. We then turn to *recency eviction* methods, wherein we propose PruLong, an end-to-end optimization method for learning which attention heads need to retain the full KV cache and which do not. PruLong saves memory while preserving long-context performance, achieving 12% smaller KV footprint than prior methods while retaining performance in challenging recall tasks. Our paper clarifies the complex tangle of long-context inference methods and paves the way for future development to minimize the KV footprint.