GEM: A Native Graph-based Index for Multi-Vector Retrieval

Summary: GEM natively indexes vector sets via set-level clustering and bridged local proximity graphs, preserving fine-grained multi-vector semantics. Metric decoupling, semantic shortcuts, multi-entry beam search, and quantized distances deliver up to 16× speedups at comparable or better accuracy. (summarized by gpt-5.6-luna on Jul 26 2026)

Paper ID: h31d8ff6884342c91
Venue: SIGMOD
Year: 2026
Pagerank: 4.9793485e-05
Overall Rank: 10,461 | 29.67%
DOI: 10.1145/3802065

Incoming Non-self Citations Over Time

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

Authors

1. Yao Tian (Hong Kong University of Science and Technology)
2. Zhoujin Tian (Hong Kong University of Science and Technology)
3. Xi Zhao (Hong Kong University of Science and Technology)
4. Ruiyuan Zhang (Hong Kong Generative AI Research & Development Center)
5. Xiaofang Zhou (Hong Kong University of Science and Technology)

BibTeX Citation

@inproceedings{tian_sigmod26,
        title = {{GEM: A Native Graph-based Index for Multi-Vector Retrieval}},
        author = {Tian, Yao and Tian, Zhoujin and Zhao, Xi and Zhang, Ruiyuan and Zhou, Xiaofang},
        series = {{SIGMOD} '26},
        booktitle = {Proceedings of the {ACM} {SIGMOD} International Conference on Management of Data},
        publisher = {Association for Computing Machinery},
        doi = {10.1145/3802065},
        url = {https://dl.acm.org/doi/10.1145/3802065},
        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 10 of 10 cited papers.

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

Rank	Cited Paper	Year	Venue	Pagerank
74	Fast Approximate Nearest Neighbor Search With The Navigating Spreading-out Graph	2019	VLDB	0.00037091678
298	Query-Aware Locality-Sensitive Hashing for Approximate Nearest Neighbor Search	2016	VLDB	0.00021833987
804	RaBitQ: Quantizing High-Dimensional Vectors with a Theoretical Error Bound for Approximate Nearest Neighbor Search	2024	SIGMOD	0.00013832333
1,328	VHP: Approximate Nearest Neighbor Search via Virtual Hypersphere Partitioning	2020	VLDB	0.00011003106
1,485	PM-LSH: A Fast and Accurate LSH Framework for High-Dimensional Approximate NN Search	2020	VLDB	0.00010523759
1,613	Starling: An I/O-Efficient Disk-Resident Graph Index Framework for High-Dimensional Vector Similarity Search on Data Segment	2024	SIGMOD	0.00010072237
1,625	Towards Efficient Index Construction and Approximate Nearest Neighbor Search in High-Dimensional Spaces	2023	VLDB	0.0001004502
2,821	Practical and Asymptotically Optimal Quantization of High-Dimensional Vectors in Euclidean Space for Approximate Nearest Neighbor Search	2025	SIGMOD	7.9711961e-05
2,969	Locality-Sensitive Hashing Scheme based on Longest Circular Co-Substring	2020	SIGMOD	7.8000797e-05
7,284	SeeSaw: Interactive Ad-hoc Search Over Image Databases	2023	SIGMOD	5.5664736e-05

Semantically Similar Papers

#	Overall Rank	Paper	Year	Venue
1	10,851	GAS: A Lightweight Framework for Filtered Search over Wide-table Vectors	2026	VLDB
2	2,060	Similarity search in the blink of an eye with compressed indices	2023	VLDB
3	8,110	Efficient and Tunable Similar Set Retrieval	2001	SIGMOD
4	13,084	Efficiency-Quality Tradeoffs for Vector Score Aggregation	2004	VLDB
5	5,982	Query-Sensitive Embeddings	2005	SIGMOD
6	10,680	Reveal Hidden Pitfalls and Navigate Next Generation of Vector Similarity Search from Task-Centric Views: [Experiments & Analysis]	2026	SIGMOD
7	13,580	VecFlow-Chamfer: A GPU-based Data Management System for High-Performance Multi-Vector Search on Superchips	2026	SIGMOD
8	2,871	Graph-Based Vector Search: An Experimental Evaluation of the State-of-the-Art	2025	SIGMOD
9	10,635	Efficient and Robust Out-Of-Distribution Vector Similarity Search with Cross-Distribution Monotonic Graph	2026	SIGMOD
10	10,462	Generalized Entity Matching with Adaptivity via Large Language Models	2026	SIGMOD