From Prefix Cache to Fusion RAG Cache: Accelerating LLM Inference in Retrieval-Augmented Generation

Summary: FusionRAG: RAG inference cache reuse beyond prefix caching. Offline chunk fusion injects cross-chunk context; online selective KV recomputation for attention-critical tokens preserves quality while cutting TTFT, outperforming prior KVCache reuse at the same recompute budget. (summarized by gpt-5-mini on Apr 11 2026)

Paper ID: 7672
Venue: SIGMOD
Year: 2026
Pagerank: 5.093636e-05
Overall Rank: 10,459 | 28.25%
DOI: 10.1145/3786655

Incoming Non-self Citations Over Time

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

Authors

1. Jiahao Wang (Approaching.AI; Hangzhou Dianzi University)
2. Weiyu Xie (Tsinghua University)
3. Mingxing Zhang (Tsinghua University)
4. Boxing Zhang (Tsinghua University)
5. Jianwei Dong (Tsinghua University)
6. Yuening Zhu (Tsinghua University)
7. Chen Lin (Tsinghua University)
8. Jingqi Tang (Approaching.AI)
9. Yaochen Han (Approaching.AI)
10. Zhiyuan Ai (Approaching.AI)
11. Xianglin Chen (Approaching.AI)
12. Yongwei Wu (Tsinghua University)
13. Congfeng Jiang (Hangzhou Dianzi University)

BibTeX Citation

@inproceedings{wang_sigmod26,
        title = {{From Prefix Cache to Fusion RAG Cache: Accelerating LLM Inference in Retrieval-Augmented Generation}},
        author = {Wang, Jiahao and Xie, Weiyu and Zhang, Mingxing and Zhang, Boxing and Dong, Jianwei and Zhu, Yuening and Lin, Chen and Tang, Jingqi and Han, Yaochen and Ai, Zhiyuan and Chen, Xianglin and Wu, Yongwei and Jiang, Congfeng},
        series = {{SIGMOD} '26},
        booktitle = {Proceedings of the {ACM} {SIGMOD} International Conference on Management of Data},
        publisher = {Association for Computing Machinery},
        doi = {10.1145/3786655},
        url = {https://dl.acm.org/doi/10.1145/3786655},
        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 1 of 1 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

Semantically Similar Papers

#	Overall Rank	Paper	Year	Venue
1	6,788	HotPrefix: Hotness-Aware KV Cache Scheduling for Efficient Prefix Sharing in LLM Inference Systems	2026	SIGMOD
2	10,916	LEGO-GraphRAG: Modularizing Graph-based Retrieval-Augmented Generation for Design Space Exploration	2025	VLDB
3	8,018	The Journey to A Knowledgeable Assistant with Retrieval-Augmented Generation (RAG)	2024	SIGMOD
4	13,309	NeutronRAG: Towards Understanding the Effectiveness of RAG from a Data Retrieval Perspective	2025	SIGMOD
5	4,430	PQCache: Product Quantization-based KVCache for Long Context LLM Inference	2025	SIGMOD
6	6,765	AquaPipe: A Quality-Aware Pipeline for Knowledge Retrieval and Large Language Models	2025	SIGMOD
7	10,357	DepCache: A KV Cache Management Framework for GraphRAG with Dependency Attention	2026	SIGMOD
8	10,290	Skyline Retrieval meets Set-Cover Chunk Merging: A Cost-Effective RAG-Sketch for Long-Context LLM QA	2026	SIGMOD
9	6,254	In-depth Analysis of Graph-based RAG in a Unified Framework	2025	VLDB
10	2,816	Cache-Craft: Managing Chunk-Caches for Efficient Retrieval-Augmented Generation	2025	SIGMOD