CoDec: Prefix-Shared Decoding Kernel for LLMs

Summary: CoDec is a dedicated decode-stage attention kernel for prefix-sharing LLM prompts, coalescing shared-prefix KV-cache accesses. It addresses tree-induced dependencies and irregular workloads by improving memory-hierarchy utilization, mitigating the context-length bottleneck. (summarized by gpt-5.6-luna on Jul 26 2026)

Paper ID: 7401
Venue: SIGMOD
Year: 2026
Pagerank: 5.093636e-05
Overall Rank: 10,210 | 29.96%
DOI: 10.1145/3802028

Incoming Non-self Citations Over Time

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

Authors

1. Zhibin Wang (Nanjing University)
2. Rui Ning (Nanjing University)
3. Chao Fang (Nanjing University)
4. Zhonghui Zhang (Nanjing University)
5. Xi Lin (Nanjing University)
6. Shaobo Ma (Nanjing University)
7. Mo Zhou (Nanjing University)
8. Xue Li (Alibaba)
9. Zhongfeng Wang (Nanjing University)
10. Chengying Huan (Nanjing University)
11. Rong Gu (Nanjing University)
12. Kun Yang (Nanjing University)
13. Guihai Chen (Nanjing University)
14. Sheng Zhong (Nanjing University)
15. Chen Tian (Nanjing University)

BibTeX Citation

@inproceedings{wang_sigmod26,
        title = {{CoDec: Prefix-Shared Decoding Kernel for LLMs}},
        author = {Wang, Zhibin and Ning, Rui and Fang, Chao and Zhang, Zhonghui and Lin, Xi and Ma, Shaobo and Zhou, Mo and Li, Xue and Wang, Zhongfeng and Huan, Chengying and Gu, Rong and Yang, Kun and Chen, Guihai and Zhong, Sheng and Tian, Chen},
        series = {{SIGMOD} '26},
        booktitle = {Proceedings of the {ACM} {SIGMOD} International Conference on Management of Data},
        publisher = {Association for Computing Machinery},
        doi = {10.1145/3802028},
        url = {https://dl.acm.org/doi/10.1145/3802028},
        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 2 of 2 cited papers.

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

Rank	Cited Paper	Year	Venue	Pagerank
2,816	Cache-Craft: Managing Chunk-Caches for Efficient Retrieval-Augmented Generation	2025	SIGMOD	8.0959781e-05
6,788	HotPrefix: Hotness-Aware KV Cache Scheduling for Efficient Prefix Sharing in LLM Inference Systems	2026	SIGMOD	5.7727874e-05

Semantically Similar Papers

#	Overall Rank	Paper	Year	Venue
1	10,432	Beluga: A CXL-Based Memory Architecture for Scalable and Efficient LLM KVCache Management	2026	SIGMOD
2	10,459	From Prefix Cache to Fusion RAG Cache: Accelerating LLM Inference in Retrieval-Augmented Generation	2026	SIGMOD
3	13,340	ContextCache: Context-Aware Semantic Cache for Multi-Turn Queries in Large Language Models	2025	VLDB
4	6,432	RetroInfer: A Vector Storage Engine for Scalable Long-Context LLM Inference	2026	VLDB
5	2,816	Cache-Craft: Managing Chunk-Caches for Efficient Retrieval-Augmented Generation	2025	SIGMOD
6	10,357	DepCache: A KV Cache Management Framework for GraphRAG with Dependency Attention	2026	SIGMOD
7	10,191	AlignedServe: Orchestrating Prefix-aware Batching to Build a High-throughput and Computing-efficient LLM Serving System	2026	SIGMOD
8	4,430	PQCache: Product Quantization-based KVCache for Long Context LLM Inference	2025	SIGMOD
9	10,259	KVDrive: A Holistic Multi-Tier KV Cache Management System for Long-Context LLM Inference	2026	SIGMOD
10	6,788	HotPrefix: Hotness-Aware KV Cache Scheduling for Efficient Prefix Sharing in LLM Inference Systems	2026	SIGMOD