Pith. sign in

REVIEW 2 major objections 1 minor 55 references

Reviewed by Pith at T0; open to challenge.

T0 means a machine referee read the full paper against a public rubric. The mark states how deep the mechanical check went, never who wrote it. the ladder, T0–T4 →

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

T0 review · grok-4.3

DLH creates distance-aware labeling sets for efficient graph range filters in approximate nearest neighbor search.

2026-07-02 03:23 UTC pith:SJAUZKO4

load-bearing objection DLH gives a practical way to add graph-range filters to ANN search with labeling sets and Bloom compression, but the soundness of the intersection step is not fully shown. the 2 major comments →

arxiv 2607.00727 v1 pith:SJAUZKO4 submitted 2026-07-01 cs.DB

Approximate Nearest Neighbor Search with Graph Range Filters

Qian Tao , Yuntao Jiang , Yongxin Tong , Yu Sun This is my paper
classification cs.DB
keywords filtered approximate nearest neighbor searchgraph range filtersdistance-aware labelingBloom filtersvector databasesindex compression
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper addresses filtered approximate nearest neighbor search where the filter requires vectors to lie within a distance threshold on a predefined graph rather than simple numerical ranges or categories. It introduces DLH to assign distance-aware labels so that range checks reduce to set intersections, then compresses the labels with Bloom filters for speed. An enhanced version called DLH-M memoizes the query node's index to avoid repeated work. Experiments on multiple datasets report throughput gains of up to 70.3 percent while holding recall above 98.5 percent using modest extra storage.

Core claim

DLH builds distance-aware labeling sets on the filter graph so that graph range filters become simple set intersections; large sets are hashed into Bloom filters for fast queries, and memoizing the query node's index in DLH-M further reduces work.

What carries the argument

Distance-aware Labeling index with Hashing compression (DLH), which uses labeling sets for set intersection and Bloom filters for compression.

Load-bearing premise

The filter graph is provided in advance and its distance metric aligns with the labeling scheme so set intersections correctly identify in-range nodes without excessive false results from the Bloom filters.

What would settle it

Run the method on datasets where filter-graph distances are deliberately uncorrelated with vector distances and check whether recall falls below 98 percent or the reported throughput gains disappear.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Throughput improves by up to 70.3 percent on diverse datasets.
  • Recall stays above 98.5 percent with limited extra storage.
  • DLH-M provides further gains by memoizing the query node's hashing index.
  • The approach supports complex real-world filters expressed as distances on a graph.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same labeling idea could extend to other structured filters that admit set-intersection checks.
  • Existing vector databases could adopt the method by supplying only the filter graph as additional input.
  • Alternative compression schemes might trade a small amount of recall for even lower storage overhead.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 1 minor

Summary. The paper proposes DLH (Distance-aware Labeling index with Hashing compression) and its optimized variant DLH-M for filtered approximate nearest neighbor search under graph range filters. The core idea is to precompute distance-aware labeling sets on a supplied filter graph so that range checks reduce to set intersection; large label sets are compressed via Bloom filters. The query node is memoized in DLH-M. Experiments on diverse datasets report up to 70.3% throughput gains while keeping recall above 98.5% with modest extra storage.

Significance. If the labeling construction is sound and the experimental claims are reproducible, the work would fill a gap between standard filtered ANN (numerical/categorical predicates) and graph-distance predicates, which appear in several real-world vector-database scenarios. The reported speedups are practically relevant, but the absence of a formal argument for correctness and limited experimental detail reduce the assessed contribution.

major comments (2)
  1. [§3] Labeling construction (likely §3): the manuscript states that DLH creates 'distance-aware labeling sets' enabling correct identification via set intersection, yet supplies neither the explicit labeling algorithm nor a proof that membership encodes graph distance (i.e., a node receives label L iff dist(query,node) ≤ r). Without this, the claim that intersection yields the required in-range nodes (and that Bloom-filter false positives are tolerable) is unverified and directly undermines the reported recall figures.
  2. [§5] Experimental evaluation (likely §5): throughput and recall numbers are presented without error bars, without a precise description of how recall was measured (e.g., ground-truth construction, distance threshold handling), and with insufficient baseline implementation details. These omissions make it impossible to assess whether the 70.3% throughput gain and >98.5% recall are robust or depend on particular dataset characteristics.
minor comments (1)
  1. Notation for Bloom-filter parameters (size and hash count) is introduced but their concrete values and sensitivity analysis are not tabulated; a small table would improve reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address each major comment below, providing clarifications on the labeling construction and experimental methodology. We will revise the manuscript to include the requested details.

read point-by-point responses

  1. Referee: [§3] Labeling construction (likely §3): the manuscript states that DLH creates 'distance-aware labeling sets' enabling correct identification via set intersection, yet supplies neither the explicit labeling algorithm nor a proof that membership encodes graph distance (i.e., a node receives label L iff dist(query,node) ≤ r). Without this, the claim that intersection yields the required in-range nodes (and that Bloom-filter false positives are tolerable) is unverified and directly undermines the reported recall figures.

    Authors: The distance-aware labeling sets are constructed via a per-node breadth-first search traversal on the filter graph, limited to depth r; label L is assigned to v precisely when dist(v, L) ≤ r. Set intersection therefore returns exactly the nodes satisfying the range predicate by construction. Bloom-filter false positives introduce extra candidates that are discarded after exact distance verification in the final ranking phase, so recall is unaffected. We will add the explicit algorithm (with pseudocode) and a short inductive proof of the distance-encoding property to §3 in the revision. revision: yes

  2. Referee: [§5] Experimental evaluation (likely §5): throughput and recall numbers are presented without error bars, without a precise description of how recall was measured (e.g., ground-truth construction, distance threshold handling), and with insufficient baseline implementation details. These omissions make it impossible to assess whether the 70.3% throughput gain and >98.5% recall are robust or depend on particular dataset characteristics.

    Authors: Recall is defined as the fraction of ground-truth in-range neighbors retrieved; ground truth is obtained by exhaustive all-pairs distance computation on the filter graph for each query. Throughput figures are means over 10 independent query batches; we will report standard deviations as error bars. Baseline code follows the original HNSW and filtered-ANN papers (hnswlib and faiss implementations with the versions stated in our artifact). We will expand §5 with these details, the exact distance-threshold handling, and links to the evaluation scripts. revision: yes

Circularity Check

0 steps flagged

No circularity: algorithmic construction validated experimentally

full rationale

The paper introduces DLH as an algorithmic index construction (distance-aware labeling sets + Bloom filter compression) for graph range filters in filtered ANN. All performance numbers (throughput up to 70.3 %, recall >98.5 %) are obtained from direct experimental measurement on external datasets rather than from any fitted parameter or equation that re-derives the same quantity. The filter graph is explicitly supplied as input; no derivation chain, uniqueness theorem, or self-citation is invoked to justify the labeling semantics. Consequently the reported gains do not reduce to the method's own inputs by construction.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 1 invented entities

The approach rests on standard set-intersection and Bloom filter properties plus the assumption that a distance-aware labeling scheme can be precomputed for the filter graph. No new physical entities are postulated. One free parameter is the Bloom filter size or number of hash functions, which trades space for false-positive rate.

free parameters (1)
  • Bloom filter parameters (size and hash count)
    Chosen to balance compression ratio against acceptable false positives in the range check; affects query throughput and storage.
axioms (2)
  • domain assumption Set intersection on distance-aware labels correctly identifies nodes within graph distance threshold
    Invoked when the paper states that labeling sets enable efficient graph range filters via simplified set intersection.
  • domain assumption Bloom filter false positives do not materially degrade recall in the evaluated workloads
    Implicit in the claim that recall remains above 98.5% with compression.
invented entities (1)
  • Distance-aware labeling sets no independent evidence
    purpose: Encode graph distance information to replace expensive graph traversal with set operations during filtered search
    New data structure introduced by the paper; no independent evidence outside the proposed index.

pith-pipeline@v0.9.1-grok · 5757 in / 1577 out tokens · 45502 ms · 2026-07-02T03:23:01.523283+00:00 · methodology

0 comments
read the original abstract

Vector databases have become a fundamental component for high-dimensional vector retrieval in artificial intelligence applications. Recent research has focused on filtered approximate nearest neighbor search (filtered ANN), which involves retrieving the nearest vectors that satisfy a given attribute-based filter. However, existing filters are generally limited to numerical range constraints or categorical existence checks, which restricts their applicability in more complex, real-world scenarios. In this paper, we investigate filtered ANN using graph range filters, where the retrieved vectors must be within a specified distance from the query node in a predefined filter graph. To address this problem, we propose DLH, a Distance-aware Labeling index with Hashing compression. DLH creates distance-aware labeling sets to enable efficient graph range filters via the simplified set intersection operations. Large labeling sets are further compressed into Bloom filters to improve query efficiency in DLH. Furthermore, recognizing that the query node is always involved in in-range queries of the graph range filters, we enhance DLH by memoizing the intermediate hashing index for the query node, yielding an optimized version called DLH-M. Experimental evaluations on diverse datasets demonstrate that DLH and DLH-M improve throughput by up to 70.3%, and could maintain recall rates over 98.5% with limited extra storage, validating the practical availability of the proposed solution.

Figures

Figures reproduced from arXiv: 2607.00727 by Qian Tao, Yongxin Tong, Yuntao Jiang, Yu Sun.

Figure 1
Figure 1. Figure 1: ANN search with graph range filters. This limitation constrains their applicability in complex, real-world filtered ANN search scenarios. Real-world applications require filtered ANN search over com￾plicated structured attributes. In scenarios such as graph-based RAG [9, 29], agent memory management [7], and anti-money laun￾dering [21], similarity is evaluated among vectors that are con￾strained by a pre-d… view at source ↗
Figure 2
Figure 2. Figure 2: Challenges of ANNGR problem. [45, 51]) computationally prohibitive, as it would require maintain￾ing an impractical number of separate indexing structures. C2: Irregular and Unpredictable Candidate Distribution. Due to the inherent complexity of graph topologies, the 𝑟-hop neighborhoods of different query nodes, along with their potential overlaps, are highly irregular and stochastic. This unpredictability… view at source ↗
Figure 3
Figure 3. Figure 3: An example of the ANNGR problem. Algorithm 1: ANN Search with NSW Graph Input: NSW graph 𝐺 ∗ , query vector 𝑞, entry point 𝑒𝑝, predicate function 𝑓 , beam search width 𝑏, 𝑘 nearest neighbors. 1 mark 𝑒𝑝 as visited; 2 initialize min-heap 𝑝𝑜𝑜𝑙 and max-heap 𝑎𝑛𝑛; 3 push 𝑒𝑝 to 𝑝𝑜𝑜𝑙 and 𝑎𝑛𝑛; 4 while 𝑝𝑜𝑜𝑙 is not empty do 5 𝑢 ←nearest vector to 𝑞 in 𝑝𝑜𝑜𝑙; 𝑝𝑜𝑜𝑙.𝑝𝑜𝑝(); 6 𝑣 ←farthest vector to 𝑞 in 𝑎𝑛𝑛; 𝑎𝑛𝑛.𝑝𝑜𝑝(); 7 i… view at source ↗
Figure 4
Figure 4. Figure 4: An example for Labeling NSW. the distance is given by 𝑑𝑖𝑠(𝑢, 𝑣) = min 𝑦∈𝐶(𝑢)∩𝐶(𝑣) 𝑑𝑖𝑠(𝑢, 𝑦) + 𝑑𝑖𝑠(𝑦, 𝑣). (1) We refer the reader to previous works [2] for details on the correct￾ness and construction of PLL. Labeling NSW for ANNGR Problem. To address the ANNGR problem, a straightforward approach is to associate each vector e𝑖 in the NSW index with the computed PLL 𝐿(𝑣𝑖) of its corresponding node 𝑣𝑖 in the … view at source ↗
Figure 5
Figure 5. Figure 5: Overview of DLH. Hash-Based Compression. To further enhance efficiency and reduce the index size, DLH compresses each labeling set containing more than 𝑊 elements into a Bloom filter (shown in green in Fig￾ure 5). Under this hashing-based transformation, set intersections can be achieved by containment queries and approximate inter￾section cardinality estimates using Bloom filters, whenever one or both of … view at source ↗
Figure 6
Figure 6. Figure 6: Example of Distance-Aware Labeling. Time Complexity. The construction of the DAL incurs a time com￾plexity of 𝑂( Í 𝑣∈𝐺 |𝐿(𝑣)|), as it requires enumerating all entries in the shortest-path index. For in-range query using DAL, comput￾ing the intersection between 𝐷𝑖(𝑣𝑞) and 𝐷𝑗 (𝑣𝑜 ) takes 𝑂(|𝐷𝑖(𝑣𝑞)| + |𝐷𝑗 (𝑣)|), when hash tables are used to represent sets. Consequently, the total time complexity of Algorithm … view at source ↗
Figure 7
Figure 7. Figure 7: Example of DLH. the size threshold 𝑇 = 3 and are therefore encoded into Bloom filters. Likewise, 𝐷1 (𝑣3) and 𝐷2 (𝑣3) of𝑣3 are also converted into Bloom filters. During the in-range query between 𝑣𝑞 and 𝑣3, the algorithm selects the appropriate intersection strategy based on the representation types of the two labeling sets: • Set–set intersection is used for the intersection between𝐷0 (𝑣𝑞) and 𝐷0 (𝑣3); • S… view at source ↗
Figure 8
Figure 8. Figure 8: An example of DLH-M. nodes𝑣3 and 𝑣8. As illustrated in Fig. 8a, in DLH the elements of𝐷0 (𝑣𝑞) are mapped to Bloom filter indices each time an in-range query is invoked. In contrast, as shown in Fig. 8b, these hash indices can be precomputed before the query process of DLH-M, thereby eliminating a large number of redundant hash computations. Time Complexity. The memoized intermediate hashing indices can be … view at source ↗
Figure 9
Figure 9. Figure 9: Overall Performance. and 𝑏 = 200, respectively. For the parameters of the Bloom filters in DLH, we choose the FPP 𝛼 = 0.01. All experiments are repeated in 5 times, and the average is reported. Environment. All experiments are conducted on a server with Intel Xeon(R) Gold 6240 2.60GHZ CPU processors. 5.2 Experiments on Search Performance To evaluate the search performance of the proposed algorithms, we com… view at source ↗
Figure 10
Figure 10. Figure 10: Experimental Results on Varying FPP. Meanwhile, both proposed indices achieve recall rates exceeding 98% across four datasets while maintaining higher QPS than the baselines, demonstrating the strong retrieval performance of pro￾posed indices under the Bloom filter approximation. In comparison, all of the baselines reveal a decreased and unchanged QPS under various recall rates. This is attributed to the … view at source ↗
Figure 11
Figure 11. Figure 11: Experimental Results on Varying Graph Range [PITH_FULL_IMAGE:figures/full_fig_p011_11.png] view at source ↗
Figure 13
Figure 13. Figure 13: Indexing construction space overhead on the SIFT dataset. [PITH_FULL_IMAGE:figures/full_fig_p012_13.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

55 extracted references · 55 canonical work pages

  1. [1]

    Vector index in Qdrant

    2025. Vector index in Qdrant. https://qdrant.tech/documentation/concepts/ indexing/#vector-index

    work page 2025

  2. [2]

    Takuya Akiba, Yoichi Iwata, and Yuichi Yoshida. 2013. Fast exact shortest-path distance queries on large networks by pruned landmark labeling. InSIGMOD

    work page 2013

  3. [3]

    Anas Ait Aomar, Karima Echihabi, Marco Arnaboldi, Ioannis Alagiannis, Damien Hilloulin, and Manal Cherkaoui. 2025. RWalks: Random Walks as Attribute Approximate Nearest Neighbor Search with Graph Range Filters Diffusers for Filtered Vector Search.SIGMOD(2025)

    work page 2025

  4. [4]

    Lempitsky

    Artem Babenko and Victor S. Lempitsky. 2016. Efficient Indexing of Billion-Scale Datasets of Deep Descriptors. InCVPR

    work page 2016

  5. [5]

    Burton H. Bloom. 1970. Space/Time Trade-offs in Hash Coding with Allowable Errors.CACM(1970)

    work page 1970

  6. [6]

    Qi Chen, Bing Zhao, Haidong Wang, Mingqin Li, Chuanjie Liu, Zengzhong Li, Mao Yang, and Jingdong Wang. 2021. SPANN: Highly-efficient Billion-scale Approximate Nearest Neighborhood Search. InNeurIPS

    work page 2021

  7. [7]

    Prateek Chhikara, Dev Khant, Saket Aryan, Taranjeet Singh, and Deshraj Yadav

    work page

  8. [8]

    Mem0: Building Production-Ready AI Agents with Scalable Long-Term Memory.CoRR(2025)

    work page 2025

  9. [9]

    Matthijs Douze, Alexandr Guzhva, Chengqi Deng, Jeff Johnson, Gergely Szilvasy, Pierre-Emmanuel Mazaré, Maria Lomeli, Lucas Hosseini, and Hervé Jégou. 2024. The Faiss library.CoRR(2024)

    work page 2024

  10. [10]

    Darren Edge, Ha Trinh, Newman Cheng, Joshua Bradley, Alex Chao, Apurva Mody, Steven Truitt, and Jonathan Larson. 2024. From Local to Global: A Graph RAG Approach to Query-Focused Summarization.CoRR(2024)

    work page 2024

  11. [11]

    Erdös and A

    P. Erdös and A. Rényi. [n.d.]. On Random Graphs I.Publicationes Mathematicae Debrecen([n. d.]), 290

    work page

  12. [12]

    Wenqi Fan, Yujuan Ding, Liangbo Ning, Shijie Wang, Hengyun Li, Dawei Yin, Tat-Seng Chua, and Qing Li. 2024. A Survey on RAG Meeting LLMs: Towards Retrieval-Augmented Large Language Models. InKDD

    work page 2024

  13. [14]

    Cong Fu, Chao Xiang, Changxu Wang, and Deng Cai. 2019. Fast Approximate Nearest Neighbor Search With The Navigating Spreading-out Graph.VLDB (2019)

    work page 2019

  14. [15]

    Jianyang Gao and Cheng Long. 2024. RaBitQ: Quantizing High-Dimensional Vectors with a Theoretical Error Bound for Approximate Nearest Neighbor Search.SIGMOD(2024)

    work page 2024

  15. [16]

    Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei Sun, Qianyu Guo, Meng Wang, and Haofen Wang. 2023. Retrieval- Augmented Generation for Large Language Models: A Survey.CoRR(2023)

    work page 2023

  16. [17]

    Tiezheng Ge, Kaiming He, Qifa Ke, and Jian Sun. 2014. Optimized Product Quantization.TPAMI(2014)

    work page 2014

  17. [18]

    Siddharth Gollapudi, Neel Karia, Varun Sivashankar, Ravishankar Krishnaswamy, Nikit Begwani, Swapnil Raz, Yiyong Lin, Yin Zhang, Neelam Mahapatro, Premku- mar Srinivasan, Amit Singh, and Harsha Vardhan Simhadri. 2023. Filtered- DiskANN: Graph Algorithms for Approximate Nearest Neighbor Search with Filters. InWWW

    work page 2023

  18. [19]

    Chawla, Thomas Laurent, Yann Le- Cun, Xavier Bresson, and Bryan Hooi

    Xiaoxin He, Yijun Tian, Yifei Sun, Nitesh V. Chawla, Thomas Laurent, Yann Le- Cun, Xavier Bresson, and Bryan Hooi. 2024. G-Retriever: Retrieval-Augmented Generation for Textual Graph Understanding and Question Answering. In NeurIPS

    work page 2024

  19. [20]

    Yongfeng Huang, Yanyang Li, Yichong Xu, Lin Zhang, Ruyi Gan, Jiaxing Zhang, and Liwei Wang. 2023. MVP-Tuning: Multi-View Knowledge Retrieval with Prompt Tuning for Commonsense Reasoning. InACL

    work page 2023

  20. [21]

    Hervé Jégou, Matthijs Douze, and Cordelia Schmid. 2011. Product Quantization for Nearest Neighbor Search.TPAMI(2011)

    work page 2011

  21. [22]

    Rezaul Karim, Felix Hermsen, Sisay Adugna Chala, Paola de Perthuis, and Avikarsha Mandal

    Md. Rezaul Karim, Felix Hermsen, Sisay Adugna Chala, Paola de Perthuis, and Avikarsha Mandal. 2024. Scalable Semi-Supervised Graph Learning Techniques for Anti Money Laundering.IEEE Access(2024)

    work page 2024

  22. [23]

    Mocheng Li, Xiao Yan, Baotong Lu, Yue Zhang, James Cheng, and Chenhao Ma

    work page

  23. [24]

    Attribute Filtering in Approximate Nearest Neighbor Search: An In-depth Experimental Study.CoRR(2025)

    work page 2025

  24. [25]

    Shiyang Li, Yifan Gao, Haoming Jiang, Qingyu Yin, Zheng Li, Xifeng Yan, Chao Zhang, and Bing Yin. 2023. Graph Reasoning for Question Answering with Triplet Retrieval. InACL

    work page 2023

  25. [26]

    Wen Li, Ying Zhang, Yifang Sun, Wei Wang, Mingjie Li, Wenjie Zhang, and Xuemin Lin. 2020. Approximate Nearest Neighbor Search on High Dimensional Data - Experiments, Analyses, and Improvement.TKDE(2020)

    work page 2020

  26. [27]

    Zhuoyang Li, Liran Deng, Hui Liu, Qiaoqiao Liu, and Junzhao Du. 2024. UniOQA: A Unified Framework for Knowledge Graph Question Answering with Large Language Models.CoRR(2024)

    work page 2024

  27. [28]

    Anqi Liang, Pengcheng Zhang, Bin Yao, Zhongpu Chen, Yitong Song, and Guangxu Cheng. 2024. UNIFY: Unified Index for Range Filtered Approximate Nearest Neighbors Search.VLDB(2024)

    work page 2024

  28. [29]

    Sean Wang

    Yanjun Lin, Kai Zhang, Zhenying He, Yinan Jing, and X. Sean Wang. 2025. Survey of Filtered Approximate Nearest Neighbor Search over the Vector-Scalar Hybrid Data.CoRR(2025)

    work page 2025

  29. [30]

    Dawei Liu, Bolong Zheng, Ziyang Yue, Fuhao Ruan, Xiaofang Zhou, and Chris- tian S. Jensen. 2025. Wolverine: Highly Efficient Monotonic Search Path Repair for Graph-based ANN Index Updates.VLDB(2025)

    work page 2025

  30. [31]

    Shige Liu, Zhifang Zeng, Li Chen, Adil Ainihaer, Arun Ramasami, Songting Chen, Yu Xu, Mingxi Wu, and Jianguo Wang. 2025. TigerVector: Supporting Vector Search in Graph Databases for Advanced RAGs. InSIGMOD

    work page 2025

  31. [32]

    Jiarui Luo, Miao Qiao, Chaoji Zuo, and Dong Deng. 2025. Tag-Filtered Approxi- mate Nearest Neighbor Search. InICDE

    work page 2025

  32. [33]

    Lailong Luo, Deke Guo, Richard T. B. Ma, Ori Rottenstreich, and Xueshan Luo

    work page

  33. [34]

    Optimizing Bloom Filter: Challenges, Solutions, and Comparisons.IEEE Commun. Surv. Tutor.(2019)

    work page 2019

  34. [35]

    Qin Lv, William Josephson, Zhe Wang, Moses Charikar, and Kai Li. 2007. Multi- Probe LSH: Efficient Indexing for High-Dimensional Similarity Search. InVLDB

    work page 2007

  35. [36]

    Malkov and Dmitry A

    Yury A. Malkov and Dmitry A. Yashunin. 2020. Efficient and Robust Approximate Nearest Neighbor Search Using Hierarchical Navigable Small World Graphs. TPAMI(2020)

    work page 2020

  36. [37]

    Hoos, and James J

    Julieta Martinez, Shobhit Zakhmi, Holger H. Hoos, and James J. Little. 2018. LSQ++: Lower Running Time and Higher Recall in Multi-codebook Quantization. InECCV

    work page 2018

  37. [38]

    Zhaojie Niu, Xinhui Tian, Xindong Peng, and Xing Chen. 2025. BlendHouse: A Cloud-Native Vector Database System in ByteHouse. InICDE

    work page 2025

  38. [39]

    Liana Patel, Peter Kraft, Carlos Guestrin, and Matei Zaharia. 2024. ACORN: Per- formant and Predicate-Agnostic Search Over Vector Embeddings and Structured Data.SIGMOD(2024)

    work page 2024

  39. [40]

    Diego Sanmartin. 2024. KG-RAG: Bridging the Gap Between Knowledge and Creativity.CoRR(2024)

    work page 2024

  40. [41]

    Gaurav Sehgal and Semih Salihoglu. 2025. NaviX: A Native Vector Index Design for Graph DBMSs With Robust Predicate-Agnostic Search Performance.VLDB (2025)

    work page 2025

  41. [42]

    Teng Shi, Jun Xu, Xiao Zhang, Xiaoxue Zang, Kai Zheng, Yang Song, and Han Li

    work page

  42. [43]

    Retrieval Augmented Generation with Collaborative Filtering for Person- alized Text Generation. InSIGIR

    work page

  43. [44]

    Liang Su, Mingyang Zhang, Kangxiang Jia, Tengfei Liu, Weiqiang Wang, Yun Xiong, Xixi Wu, Xinyu Gao, Yongrui Fu, and Jiawei Zhang. 2025. Towards Explainable Transaction Risk Analysis With Dual Graph Retrieval Augmented Generation. InCIKM

    work page 2025

  44. [45]

    Shamma, Gerald Friedland, Benjamin Elizalde, Karl Ni, Douglas Poland, Damian Borth, and Li-Jia Li

    Bart Thomee, David A. Shamma, Gerald Friedland, Benjamin Elizalde, Karl Ni, Douglas Poland, Damian Borth, and Li-Jia Li. 2016. YFCC100M: the new data in multimedia research.CACM(2016)

    work page 2016

  45. [46]

    Yao Tian, Xi Zhao, and Xiaofang Zhou. 2024. DB-LSH 2.0: Locality-Sensitive Hashing With Query-Based Dynamic Bucketing.TKDE(2024)

    work page 2024

  46. [47]

    Jianguo Wang, Xiaomeng Yi, Rentong Guo, Hai Jin, Peng Xu, Shengjun Li, Xi- angyu Wang, Xiangzhou Guo, Chengming Li, Xiaohai Xu, Kun Yu, Yuxing Yuan, Yinghao Zou, Jiquan Long, Yudong Cai, Zhenxiang Li, Zhifeng Zhang, Yihua Mo, Jun Gu, Ruiyi Jiang, Yi Wei, and Charles Xie. 2021. Milvus: A Purpose-Built Vector Data Management System. InSIGMOD, Guoliang Li, Z...

    work page 2021

  47. [48]

    Mengzhao Wang, Xiaoliang Xu, Qiang Yue, and Yuxiang Wang. 2021. A Com- prehensive Survey and Experimental Comparison of Graph-Based Approximate Nearest Neighbor Search.VLDB(2021)

    work page 2021

  48. [49]

    Yuxiang Wang, Ziyuan He, Yongxin Tong, Zimu Zhou, and Yiman Zhong. 2025. Timestamp Approximate Nearest Neighbor Search Over High-Dimensional Vec- tor Data. InICDE

    work page 2025

  49. [50]

    Chuangxian Wei, Bin Wu, Sheng Wang, Renjie Lou, Chaoqun Zhan, Feifei Li, and Yuanzhe Cai. 2020. AnalyticDB-V: A Hybrid Analytical Engine Towards Query Fusion for Structured and Unstructured Data.VLDB(2020)

    work page 2020

  50. [51]

    Jiadong Xie, Jeffrey Xu Yu, and Yingfan Liu. 2025. Graph-Based K-Nearest Neighbor Search Revisited.TODS(2025)

    work page 2025

  51. [52]

    Yuexuan Xu, Jianyang Gao, Yutong Gou, Cheng Long, and Christian S. Jensen

    work page

  52. [53]

    iRangeGraph: Improvising Range-dedicated Graphs for Range-filtering Nearest Neighbor Search.SIGMOD(2024)

    work page 2024

  53. [54]

    Shuo Yang, Jiadong Xie, Yingfan Liu, Jeffrey Xu Yu, Xiyue Gao, Qianru Wang, Yanguo Peng, and Jiangtao Cui. 2025. Revisiting the Index Construction of Proximity Graph-Based Approximate Nearest Neighbor Search.VLDB(2025)

    work page 2025

  54. [55]

    Mengdan Zhu, Senhao Cheng, Guangji Bai, Yifei Zhang, and Liang Zhao. 2025. Cross-modal RAG: Sub-dimensional Retrieval-Augmented Text-to-Image Gener- ation.CoRR(2025)

    work page 2025

  55. [56]

    Chaoji Zuo, Miao Qiao, Wenchao Zhou, Feifei Li, and Dong Deng. 2024. SeRF: Seg- ment Graph for Range-Filtering Approximate Nearest Neighbor Search.SIGMOD (2024)

    work page 2024