[2412.14838] DynamicKV: Task-Aware Adaptive KV Cache Compression for Long Context LLMs

Abstract:Efficient KV cache management in LLMs is crucial for long-context tasks like RAG and summarization. Existing KV cache compression methods enforce a fixed pattern, neglecting task-specific characteristics and reducing the retention of essential information. However, we observe distinct activation patterns across layers in various tasks, highlighting the need for adaptive strategies tailored to each task's unique demands. Based on this insight, we propose DynamicKV, a method that dynamically optimizes token retention by adjusting the number of tokens retained at each layer to adapt to the specific task. DynamicKV establishes global and per-layer maximum KV cache budgets, temporarily retaining the maximum budget for the current layer, and periodically updating the KV cache sizes of all preceding layers during inference. Our method retains only 1.7% of the KV cache size while achieving ~85% of the Full KV cache performance on LongBench. Notably, even under extreme compression (0.9%), DynamicKV surpasses state-of-the-art (SOTA) methods by 11% in the Needle-in-a-Haystack test using Mistral-7B-Instruct-v0.2. The code will be released.

DynamicKV: Task-Aware Adaptive KV Cache Compression for Long Context LLMs Xiabin Zhou, Wenbin Wang, Minyan Zeng, Jiaxian Guo, Xuebo Liu, Li Shen, Min Zhang, Liang Ding {xiabinzhou0625, liangding.liam}@gmail.com Abstract 2013), reducing…

related reading

• [2412.02252] Compressing KV Cache for Long-Context LLM Inference with Inter-Layer Attention Similarityarxiv.org
• 2407.15891arxiv.org
• Paged Attention from First Principles: A View Inside vLLM – Hamza's Bloghamzaelshafie.bearblog.dev
• Cache strategies · Hugging Facehuggingface.co
• Quest: Query-Aware Sparsity for Efficient Long-Context LLM Inferencearxiv.org
• KV Caching Explained: Optimizing Transformer Inference Efficiencyhuggingface.co
• TurboQuant: Redefining AI efficiency with extreme compressionresearch.google
• 2306.14048arxiv.org
• LLMs can invent their own compression - Rajan Agarwalrajan.sh
• Speculative KV coding: losslessly compressing KV cache by up to ~4× using a predictor modelfergusfinn.com
• [2506.06266] Cartridges: Lightweight and general-purpose long context representations via self-studyarxiv.org
• The Big LLM Architecture Comparisonmagazine.sebastianraschka.com