Efficient Cooperation-Aware Key and Value Management for LLM Inference

Summary: CoKV treats attention-head KV caching as a cooperative game, accounting for inter-head synergies rather than valuing heads independently. Precomputed contribution attribution enables globally effective cache-budget allocation for eviction and quantization, improving long-context and reasoning inference. (summarized by gpt-5.6-luna on Aug 17 2026)

Paper ID: hbbfdf1d8680832cd
Venue: VLDB
Year: 2026
Pagerank: 4.9793485e-05
Overall Rank: 10,776 | 27.55%
DOI: 10.14778/3819518.3819541

Incoming Non-self Citations Over Time

No non-self incoming citations found for this paper in this database.

Authors

1. Qiheng Sun (Hong Kong Polytechnic University; Zhejiang University)
2. Hongwei Zhang (Zhejiang University)
3. Junxu Liu (Hong Kong Polytechnic University)
4. Haocheng Xia (University of Illinois Urbana-Champaign)
5. Jinfei Liu (Zhejiang University)
6. Kui Ren (Zhejiang University)
7. Haibo Hu (Hong Kong Polytechnic University)

BibTeX Citation

@article{sun_vldb26,
        title = {{Efficient Cooperation-Aware Key and Value Management for LLM Inference}},
        author = {Sun, Qiheng and Zhang, Hongwei and Liu, Junxu and Xia, Haocheng and Liu, Jinfei and Ren, Kui and Hu, Haibo},
        journal = {PVLDB},
        series = {{VLDB} '26},
        volume = {19},
        number = {9},
        pages = {2154--2167},
        doi = {10.14778/3819518.3819541},
        url = {https://doi.org/10.14778/3819518.3819541},
        year = {2026}
}

Incoming Citations (Sorted by Pagerank)

Showing 0 of 0 citing papers.

Rank	Citing Paper	Year	Venue	Pagerank

Outgoing Citations (Sorted by Pagerank)

Showing 9 of 9 cited papers.

Citations counted here include only citations to other VLDB/SIGMOD/CIDR/PODS papers in this database.

Rank	Cited Paper	Year	Venue	Pagerank
185	CockroachDB: The Resilient Geo-Distributed SQL Database	2020	SIGMOD	0.00026072278
3,405	CoroGraph: Bridging Cache Efficiency and Work Efficiency for Graph Algorithm Execution	2024	VLDB	7.3286366e-05
4,296	Redy: Remote Dynamic Memory Cache	2022	VLDB	6.6805238e-05
4,928	Efficient Sampling Approaches to Shapley Value Approximation	2023	SIGMOD	6.3502007e-05
5,032	Put an Elephant into a Fridge: Optimizing Cache Efficiency for In-memory Key-value Stores	2020	VLDB	6.3053972e-05
5,161	HotPrefix: Hotness-Aware KV Cache Scheduling for Efficient Prefix Sharing in LLM Inference Systems	2026	SIGMOD	6.2478968e-05
6,166	Apt-Serve: Adaptive Request Scheduling on Hybrid Cache for Scalable LLM Inference Serving	2025	SIGMOD	5.8631131e-05
6,399	Catalyst: Optimizing Cache Management for Large In-memory Key-value Systems	2023	VLDB	5.798136e-05
7,935	Cache-aware load balancing of data center applications	2019	VLDB	5.4227879e-05

Semantically Similar Papers

#	Overall Rank	Paper	Year	Venue
1	13,658	ContextCache: Context-Aware Semantic Cache for Multi-Turn Queries in Large Language Models	2025	VLDB
2	10,646	From Prefix Cache to Fusion RAG Cache: Accelerating LLM Inference in Retrieval-Augmented Generation	2026	SIGMOD
3	10,407	AlignedServe: Orchestrating Prefix-aware Batching to Build a High-throughput and Computing-efficient LLM Serving System	2026	SIGMOD
4	6,560	RetroInfer: A Vector Storage Engine for Scalable Long-Context LLM Inference	2026	VLDB
5	10,426	CoDec: Prefix-Shared Decoding Kernel for LLMs	2026	SIGMOD
6	10,558	DepCache: A KV Cache Management Framework for GraphRAG with Dependency Attention	2026	SIGMOD
7	2,638	Cache-Craft: Managing Chunk-Caches for Efficient Retrieval-Augmented Generation	2025	SIGMOD
8	5,161	HotPrefix: Hotness-Aware KV Cache Scheduling for Efficient Prefix Sharing in LLM Inference Systems	2026	SIGMOD
9	4,498	PQCache: Product Quantization-based KVCache for Long Context LLM Inference	2025	SIGMOD
10	10,472	KVDrive: A Holistic Multi-Tier KV Cache Management System for Long-Context LLM Inference	2026	SIGMOD