Imagen de Portada

Not All Heads Matter: A Head-Level KV Cache Compression Method with Integrated Retrieval and Reasoning

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Fu, Yu, Cai, Zefan, Asi, Abedelkadir, Xiong, Wayne, Dong, Yue, Xiao, Wen
Formato:	Preprint
Publicado:	2024
Materias:	Computation and Language Artificial Intelligence
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

R-KV: Redundancy-aware KV Cache Compression for Reasoning Models
por: Cai, Zefan, et al.
Publicado: (2025)

PyramidKV: Dynamic KV Cache Compression based on Pyramidal Information Funneling
por: Cai, Zefan, et al.
Publicado: (2024)

Which Heads Matter for Reasoning? RL-Guided KV Cache Compression
por: Du, Wenjie, et al.
Publicado: (2025)

SwiReasoning: Switch-Thinking in Latent and Explicit for Pareto-Superior Reasoning LLMs
por: Shi, Dachuan, et al.
Publicado: (2025)

RazorAttention: Efficient KV Cache Compression Through Retrieval Heads
por: Tang, Hanlin, et al.
Publicado: (2024)

Semantic Integrity Matters: Benchmarking and Preserving High-Density Reasoning in KV Cache Compression
por: Liu, Xiang, et al.
Publicado: (2025)

CompressKV: Semantic Retrieval Heads Know What Tokens are Not Important Before Generation
por: Lin, Xiaolin, et al.
Publicado: (2025)

KV-Compress: Paged KV-Cache Compression with Variable Compression Rates per Attention Head
por: Rehg, Isaac
Publicado: (2024)

Hold Onto That Thought: Assessing KV Cache Compression On Reasoning
por: Liu, Minghui, et al.
Publicado: (2025)

HeteroCache: A Dynamic Retrieval Approach to Heterogeneous KV Cache Compression for Long-Context LLM Inference
por: Shi, Zhiyuan, et al.
Publicado: (2026)

Effectively Compress KV Heads for LLM
por: Yu, Hao, et al.
Publicado: (2024)

Lossless KV Cache Compression to 2%
por: Yang, Zhen, et al.
Publicado: (2024)

SurfaceLogicKV: Surface and Logic Attention Behaviors are All You Need for Robust KV Cache Compression
por: Li, Mengjie, et al.
Publicado: (2025)

DeltaKV: Residual-Based KV Cache Compression via Long-Range Similarity
por: Hao, Jitai, et al.
Publicado: (2026)

KVSculpt: KV Cache Compression as Distillation
por: Jiang, Bo, et al.
Publicado: (2026)

CommVQ: Commutative Vector Quantization for KV Cache Compression
por: Li, Junyan, et al.
Publicado: (2025)

OjaKV: Context-Aware Online Low-Rank KV Cache Compression
por: Zhu, Yuxuan, et al.
Publicado: (2025)

HCAttention: Extreme KV Cache Compression via Heterogeneous Attention Computing for LLMs
por: Yang, Dongquan, et al.
Publicado: (2025)

One Size Does Not Fit All: Token-Wise Adaptive Compression for KV Cache
por: Lu, Liming, et al.
Publicado: (2026)

Q-Filters: Leveraging QK Geometry for Efficient KV Cache Compression
por: Godey, Nathan, et al.
Publicado: (2025)

FreeKV: Boosting KV Cache Retrieval for Efficient LLM Inference
por: Liu, Guangda, et al.
Publicado: (2025)

Beyond RAG: Task-Aware KV Cache Compression for Comprehensive Knowledge Reasoning
por: Corallo, Giulio, et al.
Publicado: (2025)

Quantization Dominates Rank Reduction for KV-Cache Compression
por: Salfati, Samuel
Publicado: (2026)

MiniCache: KV Cache Compression in Depth Dimension for Large Language Models
por: Liu, Akide, et al.
Publicado: (2024)

ReCalKV: Low-Rank KV Cache Compression via Head Reordering and Offline Calibration
por: Yan, Xianglong, et al.
Publicado: (2025)

A Simple and Effective $L_2$ Norm-Based Strategy for KV Cache Compression
por: Devoto, Alessio, et al.
Publicado: (2024)

KeepKV: Achieving Periodic Lossless KV Cache Compression for Efficient LLM Inference
por: Tian, Yuxuan, et al.
Publicado: (2025)

ManifoldKV: Training-Free KV Cache Compression via Euclidean Outlier Detection
por: Datta, Debajyoti, et al.
Publicado: (2026)

KVReviver: Reversible KV Cache Compression with Sketch-Based Token Reconstruction
por: Yuan, Aomufei, et al.
Publicado: (2025)

RetentiveKV: State-Space Memory for Uncertainty-Aware Multimodal KV Cache Eviction
por: Liu, Sihao, et al.
Publicado: (2026)

Task-KV: Task-aware KV Cache Optimization via Semantic Differentiation of Attention Heads
por: He, Xingyang, et al.
Publicado: (2025)

Attention Is All You Need for KV Cache in Diffusion LLMs
por: Nguyen-Tri, Quan, et al.
Publicado: (2025)

EMS: Adaptive Evict-then-Merge Strategy for Head-wise KV Cache Compression Based on Global-Local Importance
por: Li, Yingxin, et al.
Publicado: (2024)

SentenceKV: Efficient LLM Inference via Sentence-Level Semantic KV Caching
por: Zhu, Yuxuan, et al.
Publicado: (2025)

Compactor: Calibrated Query-Agnostic KV Cache Compression with Approximate Leverage Scores
por: Chari, Vivek, et al.
Publicado: (2025)

Beyond Speedup -- Utilizing KV Cache for Sampling and Reasoning
por: Xing, Zeyu, et al.
Publicado: (2026)

AttentionPredictor: Temporal Patterns Matter for KV Cache Compression
por: Yang, Qingyue, et al.
Publicado: (2025)

Eigen Attention: Attention in Low-Rank Space for KV Cache Compression
por: Saxena, Utkarsh, et al.
Publicado: (2024)

LaCache: Ladder-Shaped KV Caching for Efficient Long-Context Modeling of Large Language Models
por: Shi, Dachuan, et al.
Publicado: (2025)

ZSMerge: Zero-Shot KV Cache Compression for Memory-Efficient Long-Context LLMs
por: Liu, Xin, et al.
Publicado: (2025)