# RazorAttention: Efficient KV Cache Compression Through Retrieval Heads

arXiv: https://arxiv.org/abs/2407.15891 · HTML: https://mlsanity.com/paper/2407.15891/

## Key facts

- Cited by 100 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 19 of its 28 references that are in the map.
- First posted on arXiv 22 Jul 2024.

## Placed in

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

## Builds on (most cited first, 19 of 19)

- [Efficient Streaming Language Models with Attention Sinks](https://mlsanity.com/paper/2309.17453/) (2023)
- [Generating Long Sequences with Sparse Transformers](https://mlsanity.com/paper/1904.10509/) (2019)
- [Root Mean Square Layer Normalization](https://mlsanity.com/paper/1910.07467/) (2019)
- [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)
- [Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolation](https://mlsanity.com/paper/2108.12409/) (2021)
- [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)
- [Fast Transformer Decoding: One Write-Head is All You Need](https://mlsanity.com/paper/1911.02150/) (2019)
- [Baichuan 2: Open Large-scale Language Models](https://mlsanity.com/paper/2309.10305/) (2023)
- [KVQuant: Towards 10 Million Context Length LLM Inference with KV Cache Quantization](https://mlsanity.com/paper/2401.18079/) (2024)
- [FlexGen: High-Throughput Generative Inference of Large Language Models with a Single GPU](https://mlsanity.com/paper/2303.06865/) (2023)
- [Scissorhands: Exploiting the Persistence of Importance Hypothesis for LLM KV Cache Compression at Test Time](https://mlsanity.com/paper/2305.17118/) (2023)
- [LLM-QAT: Data-Free Quantization Aware Training for Large Language Models](https://mlsanity.com/paper/2305.17888/) (2023)
- [Griffin: Mixing Gated Linear Recurrences with Local Attention for Efficient Language Models](https://mlsanity.com/paper/2402.19427/) (2024)
- [Leave No Context Behind: Efficient Infinite Context Transformers with Infini-attention](https://mlsanity.com/paper/2404.07143/) (2024)
- [Data Engineering for Scaling Language Models to 128K Context](https://mlsanity.com/paper/2402.10171/) (2024)
- [Outlier Suppression: Pushing the Limit of Low-bit Transformer Language Models](https://mlsanity.com/paper/2209.13325/) (2022)
- [PyramidInfer: Pyramid KV Cache Compression for High-throughput LLM Inference](https://mlsanity.com/paper/2405.12532/) (2024)

## Built on by (most cited first, 30 of 100)

- [DuoAttention: Efficient Long-Context LLM Inference with Retrieval and Streaming Heads](https://mlsanity.com/paper/2410.10819/) (2024)
- [ShadowKV: KV Cache in Shadows for High-Throughput Long-Context LLM Inference](https://mlsanity.com/paper/2410.21465/) (2024)
- [A Survey on Large Language Model Acceleration based on KV Cache Management](https://mlsanity.com/paper/2412.19442/) (2024)
- [Keep the Cost Down: A Review on Methods to Optimize LLM' s KV-Cache Consumption](https://mlsanity.com/paper/2407.18003/) (2024)
- [Not All Heads Matter: A Head-Level KV Cache Compression Method with Integrated Retrieval and Reasoning](https://mlsanity.com/paper/2410.19258/) (2024)
- [Attention Heads of Large Language Models: A Survey](https://mlsanity.com/paper/2409.03752/) (2024)
- [InfiniPot-V: Memory-Constrained KV Cache Compression for Streaming Video Understanding](https://mlsanity.com/paper/2506.15745/) (2025)
- [Twilight: Adaptive Attention Sparsity with Hierarchical Top-p Pruning](https://mlsanity.com/paper/2502.02770/) (2025)
- [RocketKV: Accelerating Long-Context LLM Inference via Two-Stage KV Cache Compression](https://mlsanity.com/paper/2502.14051/) (2025)
- [KVTuner: Sensitivity-Aware Layer-Wise Mixed-Precision KV Cache Quantization for Efficient and Nearly Lossless LLM Inference](https://mlsanity.com/paper/2502.04420/) (2025)
- [TokenSelect: Efficient Long-Context Inference and Length Extrapolation for LLMs via Dynamic Token-Level KV Cache Selection](https://mlsanity.com/paper/2411.02886/) (2024)
- [Position: Episodic Memory is the Missing Piece for Long-Term LLM Agents](https://mlsanity.com/paper/2502.06975/) (2025)
- [Cognitive Memory in Large Language Models](https://mlsanity.com/paper/2504.02441/) (2025)
- [Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Models](https://mlsanity.com/paper/2601.14004/) (2026)
- [InstInfer: In-Storage Attention Offloading for Cost-Effective Long-Context LLM Inference](https://mlsanity.com/paper/2409.04992/) (2024)
- [Compactor: Calibrated Query-Agnostic KV Cache Compression with Approximate Leverage Scores](https://mlsanity.com/paper/2507.08143/) (2025)
- [Past- and Future-Informed KV Cache Policy with Salience Estimation in Autoregressive Video Diffusion](https://mlsanity.com/paper/2601.21896/) (2026)
- [Intrinsic Entropy of Context Length Scaling in LLMs](https://mlsanity.com/paper/2502.01481/) (2025)
- [Dialogue Injection Attack: Jailbreaking LLMs through Context Manipulation](https://mlsanity.com/paper/2503.08195/) (2025)
- [FreeKV: Boosting KV Cache Retrieval for Efficient LLM Inference](https://mlsanity.com/paper/2505.13109/) (2025)
- [LightTransfer: Your Long-Context LLM is Secretly a Hybrid Model with Effortless Adaptation](https://mlsanity.com/paper/2410.13846/) (2024)
- [Efficient Prompt Compression with Evaluator Heads for Long-Context Transformer Inference](https://mlsanity.com/paper/2501.12959/) (2025)
- [KV Cache Compression for Inference Efficiency in LLMs: A Review](https://mlsanity.com/paper/2508.06297/) (2025)
- [MEPIC: Memory Efficient Position Independent Caching for LLM Serving](https://mlsanity.com/paper/2512.16822/) (2025)
- [More Tokens, Lower Precision: Towards the Optimal Token-Precision Trade-off in KV Cache Compression](https://mlsanity.com/paper/2412.12706/) (2024)
- [Which Heads Matter for Reasoning? RL-Guided KV Cache Compression](https://mlsanity.com/paper/2510.08525/) (2025)
- [A Little Goes a Long Way: Efficient Long Context Training and Inference with Partial Contexts](https://mlsanity.com/paper/2410.01485/) (2024)
- [HeadInfer: Memory-Efficient LLM Inference by Head-wise Offloading](https://mlsanity.com/paper/2502.12574/) (2025)
- [ReST-KV: Robust KV Cache Eviction with Layer-wise Output Reconstruction and Spatial-Temporal Smoothing](https://mlsanity.com/paper/2605.08840/) (2026)
- [Rethinking Key-Value Cache Compression Techniques for Large Language Model Serving](https://mlsanity.com/paper/2503.24000/) (2025)

## Abstract (arXiv)

The memory and computational demands of Key-Value (KV) cache present significant challenges for deploying long-context language models. Previous approaches attempt to mitigate this issue by selectively dropping tokens, which irreversibly erases critical information that might be needed for future queries. In this paper, we propose a novel compression technique for KV cache that preserves all token information. Our investigation reveals that: i) Most attention heads primarily focus on the local context; ii) Only a few heads, denoted as retrieval heads, can essentially pay attention to all input tokens. These key observations motivate us to use separate caching strategy for attention heads. Therefore, we propose RazorAttention, a training-free KV cache compression algorithm, which maintains a full cache for these crucial retrieval heads and discards the remote tokens in non-retrieval heads. Furthermore, we introduce a novel mechanism involving a "compensation token" to further recover the information in the dropped tokens. Extensive evaluations across a diverse set of large language models (LLMs) demonstrate that RazorAttention achieves a reduction in KV cache size by over 70% without noticeable impacts on performance. Additionally, RazorAttention is compatible with FlashAttention, rendering it an efficient and plug-and-play solution that enhances LLM inference efficiency without overhead or retraining of the original model.