paper-with-me

홈 › Papers

KV Packet: Recomputation-Free Context-Independent KV Caching for LLMs

2026-04-14 · Chuangtao Chen, Grace Li Zhang, Xunzhao Yin, Cheng Zhuo, Bing Li, Ulf Schlichtmann arxiv

Large Language Models (LLMs) rely heavily on Key-Value (KV) caching to minimize inference latency. However, standard KV caches are context-dependent: reusing a cached document in a new context requires recomputing KV states to account for shifts in attention distribution. Existing solutions such as CacheBlend, EPIC, and SAM-KV mitigate this issue by selectively recomputing a subset of tokens; however, they still incur non-negligible computational overhead (FLOPs) and increased Time-to-First-Token (TTFT) latency. In this paper, we propose KV Packet, a recomputation-free cache reuse framework that treats cached documents as immutable ``packets'' wrapped in light-weight trainable soft-token adapters, which are trained via self-supervised distillation to bridge context discontinuities. Experiments on Llama-3.1 and Qwen2.5 demonstrate that the proposed KV Packet method achieves near-zero FLOPs and lower TTFT than recomputation-based baselines, while retaining F1 scores comparable to those of the full recomputation baseline.

📄 PDF Abstract BibTeX arXiv:2604.13226

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

MPIC: Position-Independent Multimodal Context Caching System for Efficient MLLM Serving

2025-02-04 · Shiju Zhao, Junhao Hu, Rongxiao Huang, Jiaqi Zheng 외

The context caching technique is employed to accelerate the Multimodal Large Language Model (MLLM) inference by prevailing serving platforms currently. However, this approach merely reuses the Key-Value (KV) cache of the…

Language ModelingLanguage ModellingLarge Language ModelMultimodal Large Language Model+2

EntropyCache: Decoded Token Entropy Guided KV Caching for Diffusion Language Models

2026-03-19 · Minsoo Cheong, Donghyun Son, Woosang Lim, Sungjoo Yoo arxiv

Diffusion-based large language models (dLLMs) rely on bidirectional attention, which prevents lossless KV caching and requires a full forward pass at every denoising step. Existing approximate KV caching methods reduce t…

Understanding Is Done Early: A Depth Division of Labor in Large Language Models and Its Use for Unbounded-Context Memory

2026-07-30 · Hanzuo Liu, Xuan Qi, Chunyu Liu, Haotian Zhong 외 arxiv

Transformer depth is not used uniformly: lower and middle layers build semantic representations, while upper layers increasingly specialize them for prediction. We turn this division of labor into CoMem (Comprehension Me…

MEPIC: Memory Efficient Position Independent Caching for LLM Serving

2025-12-18 · Qian Wang, Zahra Yousefijamarani, Morgan Lindsay Heisler, Rongzhi Gu 외 arxiv

Modern LLM applications such as deep-research assistants, coding agents, and Retrieval-Augmented Generation (RAG) systems, repeatedly process long prompt histories containing shared document or code chunks, creating sign…

SemPIC: Learning Semantic Position-Independent KV Caches

2026-07-30 · Hui Xie, Peng Xiao, Yutong Deng\textsuperscript, Shuoran Dou 외 arxiv

Long-context retrieval and agentic workloads repeatedly reuse the same documents under changing instructions, histories, and document orders. Prefix caching cannot exploit this reuse, while position-independent caching (…