mlsanityData to 7 Oct 2026

FreqKV: Key-Value Compression in Frequency Domain for Context Window Extension

arXiv 2505.00570 ↗First posted 1 May 2025cs.CL cs.AIcited by 7

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

Abstract (arXiv)

Existing key-value (KV) cache compression methods for large language models (LLMs) often rely on token eviction, which risks losing critical local information in both long prefilling and decoding scenarios. When extrapolating beyond the pretrained context length, their performance degrades sharply on long-context benchmarks. Motivated by the observation in the frequency domain that the context information is concentrated in the low-frequency components, we propose FreqKV, a parameter-free and architecture-agnostic approach. It iteratively compresses the increasing KV cache in the frequency domain, allowing models to process lengthy contexts efficiently. With minimal training at 8K length, FreqKV extends the context window of LLaMA-2-7B up to 256K tokens while maintaining stable perplexity. Extensive experiments across prefilling and decoding demonstrate that FreqKV enables robust context window extension and consistently outperforms existing KV cache compression methods on LLaMA-2 and LLaMA-3, highlighting its effectiveness for both understanding and generation in long contexts.

Key facts. Cited by 7 papers in the map (data through 7 Oct 2026). Placed in KV cache optimization, one of 12 directions in Efficient LLM inference and serving. Builds on 30 of its 37 references that are in the map. First posted on arXiv 1 May 2025.

Builds on

30 of 37 references.

Closest ancestors · 8

Same direction, newest first

  1. 2025FastKV: Decoupling of Context Reduction and KV Cache Compression for Prefill-Decoding Acceleration3 Feb
  2. 2024Discovering the Gems in Early Layers: Accelerating Long-Context LLMs with 1000x Input Token Reduction25 Sep
  3. 2024ThinK: Thinner Key Cache by Query-Driven Pruning30 Jul
  4. 2024Model Tells You Where to Merge: Adaptive KV Cache Merging for LLMs on Long-Context Tasks11 Jul
  5. 2024D2O: Dynamic Discriminative Operations for Efficient Long-Context Inference of Large Language Models18 Jun
  6. 2024LoCoCo: Dropping In Convolutions for Long Context Compression8 Jun
  7. 2024PyramidKV: Dynamic KV Cache Compression based on Pyramidal Information Funneling4 Jun
  8. 2024SnapKV: LLM Knows What You are Looking for Before Generation22 Apr

Other references · 22

  1. 2021LoRA: Low-Rank Adaptation of Large Language Modelscited by 14,084
  2. 2021RoFormer: Enhanced Transformer with Rotary Position Embeddingcited by 4,581
  3. 2020Longformer: The Long-Document Transformercited by 3,362
  4. 2023FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioningcited by 2,551
  5. 2023Efficient Streaming Language Models with Attention Sinkscited by 2,115
  6. 2023LongBench: A Bilingual, Multitask Benchmark for Long Context Understandingcited by 1,378
  7. 2024RULER: What's the Real Context Size of Your Long-Context Language Models?cited by 1,270
  8. 2021Train Short, Test Long: Attention with Linear Biases Enables Input Length Extrapolationcited by 1,085
  9. 2015Teaching Machines to Read and Comprehendcited by 1,039
  10. 2023Extending Context Window of Large Language Models via Positional Interpolationcited by 721
  11. 2019Compressive Transformers for Long-Range Sequence Modellingcited by 699
  12. 2024MInference 1.0: Accelerating Pre-filling for Long-Context LLMs via Dynamic Sparse Attentioncited by 375
10 more
  1. 2019Frequency Principle: Fourier Analysis Sheds Light on Deep Neural Networkscited by 360
  2. 2024LongRoPE: Extending LLM Context Window Beyond 2 Million Tokenscited by 302
  3. 2023Landmark Attention: Random-Access Infinite Context Length for Transformerscited by 206
  4. 2020Learning in the Frequency Domaincited by 196
  5. 2024InfLLM: Training-Free Long-Context Extrapolation for LLMs with an Efficient Context Memorycited by 192
  6. 2023LM-Infinite: Zero-Shot Extreme Length Generalization for Large Language Modelscited by 120
  7. 2024Long Context Compression with Activation Beaconcited by 101
  8. 2019Theory of the Frequency Principle for General Deep Neural Networkscited by 45
  9. 2024LongGenBench: Long-context Generation Benchmarkcited by 27
  10. 2023Fourier Transformer: Fast Long Range Modeling by Removing Sequence Redundancy with FFT Operatorcited by 19

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 2; 2026: 5.

7 papers.

  1. 2025FAEDKV: Infinite-Window Fourier Transform for Unbiased KV Cache Compressioncited by 11
  2. 2025Fourier Compressor: Frequency-Domain Visual Token Compression for Vision-Language Modelscited by 3
  3. 2026Adaptive KV Cache Reuse for Fast Long-Context LLM Servingcited by 2
  4. 2026Information-Aware KV Cache Compression for Long Reasoningcited by 2
  5. 2026UT-ACA: Uncertainty-Triggered Adaptive Context Allocation for Long-Context Inferencecited by 0
  6. 2026Fre-Res: Frequency-Residual Video Token Compression for Efficient Video MLLMscited by 0
  7. 2026Codec-Gauge: Learning Compression-Friendly Gauges for Transformer KV Cachescited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2025Key, Value, Compress: A Systematic Exploration of KV Cache Compression Techniques
  2. 2026PuzzleKV: Page-Wise Low-Rank Decomposition for KV Cache Compression
  3. 2025Homogeneous Keys, Heterogeneous Values: Exploiting Local KV Cache Asymmetry for Long-Context LLMs
  4. 2026CompressKV: Semantic-Retrieval-Guided KV-Cache Compression for Resource-Efficient Long-Context LLM Inference
  5. 2026GRKV: Global Regression for Training-Free KV Cache Compression in Long-Context LLMs
  6. 2025TreeKV: Smooth Key-Value Cache Compression with Tree Structures
  7. 2025Cache Me If You Must: Adaptive Key-Value Quantization for Large Language Models
  8. 2025TailorKV: A Hybrid Framework for Long-Context Inference via Tailored KV Cache Optimization