pith. sign in

arxiv: 2605.17992 · v1 · pith:S5BJGEIPnew · submitted 2026-05-18 · 💻 cs.OS · cs.DB

PipeANN-Filter: An Efficient Filtered Vector Search System on SSD

Hao Guo , Jiwu Shu , Youyou Lu This is my paper

Pith reviewed 2026-05-20 00:39 UTC · model grok-4.3

classification 💻 cs.OS cs.DB
keywords filtered vector searchSSDprobabilistic data structuresBloom filtersapproximate nearest neighborI/O optimizationattribute filtering
0
0 comments X

The pith

PipeANN-Filter explores a superset of valid vectors to reduce SSD I/O in filtered vector searches.

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

This paper introduces PipeANN-Filter for filtered vector search on solid-state drives. Traditional systems explore only vectors that meet attribute constraints, requiring many SSD reads for attribute data during search. PipeANN-Filter instead explores a larger superset of vectors likely to satisfy the constraints and uses probabilistic structures to select them, verifying attributes only on the final top-k results. The approach accepts a few extra vector explorations to avoid most attribute reads from the drive. Readers would care because vector search with filters is central to recommendation and retrieval systems, and lowering I/O costs on common storage hardware can improve query speed and scalability.

Core claim

PipeANN-Filter explores a superset of valid vectors, and performs attribute verification after getting the top-k closest result vectors. This allows PipeANN-Filter to leverage probabilistic data structures (e.g., Bloom filters) to identify the superset, trading off a small number of false-positive vector explorations for a massive reduction in SSD I/O for attribute reading.

What carries the argument

Superset exploration via probabilistic data structures to identify candidate vectors before attribute verification, which defers and minimizes SSD I/O.

If this is right

  • Search latency drops because far fewer attribute values are read from SSD during the process.
  • Throughput rises in filtered search tasks since I/O overhead falls while result quality stays intact.
  • The system scales better on standard SSD hardware for combined similarity and attribute queries.
  • Performance gains hold when the filter is selective enough to keep the superset size modest.

Where Pith is reading between the lines

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

  • The same deferral of verification could apply to other I/O-heavy search tasks that mix similarity with constraints.
  • Faster on-device computation would widen the range of filter selectivities where the tradeoff stays favorable.
  • Integrating the method with data layout techniques might further cut the remaining I/O cost.

Load-bearing premise

The time saved by avoiding most attribute reads from the SSD must exceed the added time for exploring extra vectors and running probabilistic checks.

What would settle it

A workload test with low attribute filter selectivity where the number of false-positive explorations grows large enough to increase overall latency above that of baseline systems.

Figures

Figures reproduced from arXiv: 2605.17992 by Hao Guo, Jiwu Shu, Youyou Lu.

Figure 1
Figure 1. Figure 1: Overview of an on-SSD graph-based ANNS index. (a) On-SSD data layout. Each record stores a full-precision vector and its neighbor IDs. (b) Vector access pattern during a search. Records of vectors along the search path are fetched from the SSD. Their neighbors’ PQ-compressed vectors are accessed in memory for distance comparison, without in￾volving the SSD. Other vectors are not accessed. • We design PipeA… view at source ↗
Figure 2
Figure 2. Figure 2: Search throughput of different filtering mech￾anisms across varying selectivities. Dataset: LAION100M. Target recall: 0.9. The blue line shows the performance of our system, PipeANN-Filter. dataset [18], 500 million items contain ∼300GB attributes (e.g., product features and reviews). Therefore, to support large-scale vectors with attributes, it is crucial to store both vectors and their attributes on SSDs… view at source ↗
Figure 3
Figure 3. Figure 3: Comparison of speculative pre-/in-filtering with strict pre-/in-filtering. 9 3 1 2 5 6 8 0 7 4 SSD Record 𝑉𝑉8 nbrs attrs 2-hop nbrs Vector Index 0 1 2 3 … 8 9 PQ-compressed vectors DRAM Attribute Index Label-Filter (§4.3.1) 1 3 5 2 4 2 6 7 ID Label 0 1 2 0 1 4 0 5 2 Val ID Range-Filter (§4.3.2) Label Count Bloom Filter 3 2 3 … Histogram 0 5 9 … Quant Value 0 2 … Cost-Estimation (§4.2) [PITH_FULL_IMAGE:fig… view at source ↗
Figure 4
Figure 4. Figure 4: PipeANN-Filter overview. 4 PipeANN-Filter Design and Implementation We design and implement PipeANN-Filter, a filtered ANNS system on SSD. To build a system atop speculative filtering, PipeANN-Filter tackles two main design challenges: C1: False-positive-aware cost estimation. In-memory fil￾tered ANNS systems [25] directly use query selectivity to estimate query costs and choose filtering mechanisms (e.g.,… view at source ↗
Figure 5
Figure 5. Figure 5: Search throughput on YT5M and YFCC10M. 5.2 Overall Performance In this section, we evaluate PipeANN-Filter on two label￾filtering datasets: YT5M and YFCC10M. YT5M evaluates label OR conditions, while YFCC10M evaluates label AND conditions. We compare PipeANN-Filter against PipeANN￾BaseFilter and Milvus. Throughput [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Search latency on YT5M and YFCC10M. throughput of PipeANN-BaseFilter on YT5M (at 0.95 recall) and YFCC10M (at 0.99 recall). This gap stems from the rapidly increasing cost of post-filtering. While both in-filtering and post-filtering require linearly more I/O to achieve higher accuracy (with a larger 𝐿), post-filtering’s cost grows at a much steeper rate (as shown in [PITH_FULL_IMAGE:figures/full_fig_p010… view at source ↗
Figure 8
Figure 8. Figure 8: Search latency on LAION100M. Throughput (Op/s) (a) LabelOr PipeANN-Filter PipeANN-BaseFilter (b) Range (c) Hybrid Recall10@10 0 5k 10k 0.8 0.9 1.0 0.8 0.9 1.0 0.8 0.9 1.0 [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Search throughput on LAION100M. where PipeANN-BaseFilter fails to reach a high recall within a 10ms latency scale. This shows that post-filtering some￾times struggles to find enough valid nearby vectors under tight range constraints. In contrast, PipeANN-Filter’s spec￾ulative in-filtering maintains graph connectivity, delivering both superior recall and throughput. 5.4 In-Depth Analysis In this section, we… view at source ↗
read the original abstract

We propose PipeANN-Filter, an efficient filtered vector search system on SSD. Unlike existing systems that explore only valid vectors (i.e., those satisfying the attribute constraints) during search, PipeANN-Filter explores a superset of valid vectors, and performs attribute verification after getting the top-k closest result vectors. This allows PipeANN-Filter to leverage probabilistic data structures (e.g., Bloom filters) to identify the superset, trading off a small number of false-positive vector explorations for a massive reduction in SSD I/O for attribute reading. Evaluations show that PipeANN-Filter improves search latency and throughput compared to state-of-the-art systems. PipeANN-Filter is open-source at https://github.com/thustorage/PipeANN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 1 minor

Summary. The manuscript proposes PipeANN-Filter, a filtered vector search system for SSD storage. Unlike prior systems that restrict search to only attribute-valid vectors, PipeANN-Filter identifies a probabilistic superset of valid vectors via structures such as Bloom filters, retrieves the top-k closest vectors from this superset, and performs attribute verification afterward. This design trades a controlled number of false-positive vector explorations for substantially reduced SSD I/O on attribute data. The paper reports that the resulting system improves search latency and throughput relative to state-of-the-art baselines and releases the implementation as open source.

Significance. If the reported latency and throughput gains are reproducible across realistic selectivities and data distributions, the work would offer a practical engineering contribution to vector search on secondary storage by demonstrating that modest extra computation can yield large I/O savings when attribute filtering dominates cost.

major comments (2)
  1. [Evaluation] Evaluation section: the abstract asserts latency and throughput improvements but supplies no quantitative results, error bars, workload characteristics, selectivity ranges, or direct comparison numbers against baselines. Without these data the central performance claim cannot be verified and the tradeoff between extra vector explorations and I/O savings remains unquantified.
  2. [Design] Design and Bloom-filter integration: the claim that I/O savings from the probabilistic superset outweigh the cost of false-positive explorations is load-bearing, yet no measured false-positive rates, ablation of the probabilistic component, or sensitivity analysis across filter selectivities are referenced. If moderate selectivity or vector-data-dominant workloads are present, the net gain may disappear.
minor comments (1)
  1. [Abstract] The abstract and introduction would benefit from a brief statement of the target workload assumptions (e.g., typical filter selectivity and vector dimensionality) to help readers assess applicability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback on PipeANN-Filter. We address each major comment below and will revise the manuscript accordingly to improve clarity and completeness of the evaluation and design claims.

read point-by-point responses

  1. Referee: [Evaluation] Evaluation section: the abstract asserts latency and throughput improvements but supplies no quantitative results, error bars, workload characteristics, selectivity ranges, or direct comparison numbers against baselines. Without these data the central performance claim cannot be verified and the tradeoff between extra vector explorations and I/O savings remains unquantified.

    Authors: We agree that the abstract would be strengthened by including specific quantitative results. The evaluation section presents latency and throughput comparisons against baselines, but to make the central claims immediately verifiable, we will revise the abstract to report key numbers (e.g., latency reductions and throughput gains) along with the tested selectivity ranges and workload characteristics. We will also ensure error bars and direct comparison tables are clearly highlighted in the revised evaluation section. revision: yes

  2. Referee: [Design] Design and Bloom-filter integration: the claim that I/O savings from the probabilistic superset outweigh the cost of false-positive explorations is load-bearing, yet no measured false-positive rates, ablation of the probabilistic component, or sensitivity analysis across filter selectivities are referenced. If moderate selectivity or vector-data-dominant workloads are present, the net gain may disappear.

    Authors: This is a valid point. The current manuscript emphasizes end-to-end performance but does not explicitly report false-positive rates or include an ablation isolating the probabilistic filter. In the revision we will add measured false-positive rates for the Bloom filters, an ablation study removing the probabilistic component, and sensitivity analysis across a broader range of selectivities (including moderate values). We will also discuss scenarios where attribute filtering is not the dominant cost and note conditions under which net gains may be limited. revision: yes

Circularity Check

0 steps flagged

No circularity: engineering design with no derivation chain

full rationale

The paper presents PipeANN-Filter as a new systems architecture that uses Bloom filters to identify a probabilistic superset of vectors before attribute verification. No equations, fitted parameters, predictions, or first-principles derivations appear in the provided text or abstract. The approach is described as an engineering tradeoff trading limited extra vector explorations for reduced attribute I/O. Claims rest on implementation and evaluation rather than any self-referential reduction or self-citation that bears the central load. The design is self-contained and externally falsifiable via open-source code and benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard systems assumptions about SSD I/O being the dominant cost and on the effectiveness of Bloom filters for superset identification; no free parameters or invented entities are introduced in the abstract.

axioms (2)
  • domain assumption SSD random I/O for attribute reads is the primary performance bottleneck in filtered vector search.
    Invoked implicitly when the design trades extra vector explorations for reduced attribute reads.
  • domain assumption Probabilistic data structures can accurately identify a useful superset of attribute-matching vectors with low false-positive overhead.
    Core premise of the PipeANN-Filter approach described in the abstract.

pith-pipeline@v0.9.0 · 5650 in / 1368 out tokens · 39792 ms · 2026-05-20T00:39:54.979265+00:00 · methodology

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

54 extracted references · 54 canonical work pages · 6 internal anchors

  1. [1]

    Sami Abu-El-Haija, Nisarg Kothari, Joonseok Lee, Paul Natsev, George Toderici, Balakrishnan Varadarajan, and Sudheendra Vijaya- narasimhan. 2016. YouTube-8M: A Large-Scale Video Classifica- tion Benchmark.CoRRabs/1609.08675 (2016). arXiv:1609.08675 http://arxiv.org/abs/1609.08675

    work page internal anchor Pith review Pith/arXiv arXiv 2016

  2. [2]

    Artem Babenko and Victor Lempitsky. 2012. The inverted multi-index. In2012 IEEE Conference on Computer Vision and Pattern Recognition. 3069–3076. doi:10.1109/CVPR.2012.6248038

    work page doi:10.1109/cvpr.2012.6248038 2012

  3. [3]

    Dmitry Baranchuk, Artem Babenko, and Yury Malkov. 2018. Revisiting the inverted indices for billion-scale approximate nearest neighbors. InProceedings of the European Conference on Computer Vision (ECCV). 202–216

    work page 2018

  4. [4]

    Burton H. Bloom. 1970. Space/time trade-offs in hash coding with allowable errors.Communication of the ACM13, 7 (1970), 422–426. doi:10.1145/362686.362692

    work page doi:10.1145/362686.362692 1970

  5. [5]

    Yuzheng Cai, Jiayang Shi, Yizhuo Chen, and Weiguo Zheng. 2024. Navi- gating Labels and Vectors: A Unified Approach to Filtered Approximate Nearest Neighbor Search. , Article 246 (2024). doi:10.1145/3698822

    work page doi:10.1145/3698822 2024

  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 neighbor search. InPro- ceedings of the 35th International Conference on Neural Information Processing Systems (NIPS ’21). Curran Associates Inc., Red Hook, NY, USA, Article 398

    work page 2021

  7. [7]

    Weijian Chen, Haotian Liu, Yangshen Deng, Long Xiang, Liang Huang, Gezi Li, and Bo Tang. 2026. AlayaLaser: Efficient Index Layout and Search Strategy for Large-scale High-dimensional Vector Similarity Search. arXiv:2602.23342 [cs.DB]https://arxiv.org/abs/2602.23342

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  8. [8]

    Xiaoyu Chen, Jinxiu Qu, Yitong Song, Shuhang Lu, Huiling Li, Minghui Jiang, Wei Zhou, Jianliang Xu, Xuanhe Zhou, and Fan Wu

    work page

  9. [9]

    arXiv:2603.01779 [cs.DB]https://arxiv.org/abs/2603.01779

    Disk-Resident Graph ANN Search: An Experimental Evaluation. arXiv:2603.01779 [cs.DB]https://arxiv.org/abs/2603.01779

    work page arXiv

  10. [10]

    Matthijs Douze, Alexandr Guzhva, Chengqi Deng, Jeff Johnson, Gergely Szilvasy, Pierre-Emmanuel Mazaré, Maria Lomeli, Lucas Hosseini, and Hervé Jégou. 2024. The Faiss library. (2024). arXiv:2401.08281 [cs.LG]

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  11. [11]

    Facebook. 2026. RocksDB: A Persistent Key-Value Store for Flash and RAM Storage.http://rocksdb.org/

    work page 2026

  12. [12]

    Andersen, Michael Kaminsky, and Michael D

    Bin Fan, Dave G. Andersen, Michael Kaminsky, and Michael D. Mitzen- macher. 2014. Cuckoo Filter: Practically Better Than Bloom. InPro- ceedings of the 10th ACM International on Conference on Emerging Networking Experiments and Technologies (CoNEXT ’14). Association for Computing Machinery, Sydney, Australia, 75–88. doi:10.1145/2674 005.2674994

    work page doi:10.1145/2674 2014

  13. [13]

    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. InProceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD ’24). Association for Computing Machinery, Barcelona, Spain, 6491–6501. doi:...

    work page doi:10.1145/3637528.3671470 2024

  14. [14]

    Cong Fu, Chao Xiang, Changxu Wang, and Deng Cai. 2019. Fast approximate nearest neighbor search with the navigating spreading- out graph. InProceedings of the VLDB Endowment (VLDB ’19). VLDB Endowment, Los Angeles, CA, USA, 461–474. doi:10.14778/3303753.3 303754

    work page doi:10.14778/3303753.3 2019

  15. [15]

    Tiezheng Ge, Kaiming He, Qifa Ke, and Jian Sun. 2013. Optimized Product Quantization for Approximate Nearest Neighbor Search. In 2013 IEEE Conference on Computer Vision and Pattern Recognition. 2946–

    work page 2013

  16. [16]

    doi:10.1109/CVPR.2013.379

    work page doi:10.1109/cvpr.2013.379 2013

  17. [17]

    Siddharth Gollapudi, Neel Karia, Varun Sivashankar, Ravishankar Kr- ishnaswamy, Nikit Begwani, Swapnil Raz, Yiyong Lin, Yin Zhang, Neelam Mahapatro, Premkumar Srinivasan, Amit Singh, and Har- sha Vardhan Simhadri. 2023. Filtered-DiskANN: Graph Algorithms for Approximate Nearest Neighbor Search with Filters. InProceedings of the ACM Web Conference 2023 (WW...

    work page doi:10.1145/3543507.3583552 2023

  18. [18]

    Martin Grohe. 2020. word2vec, node2vec, graph2vec, X2vec: Towards a Theory of Vector Embeddings of Structured Data. InProceedings of the 39th ACM SIGMOD-SIGACT-SIGAI Symposium on Principles of Database Systems(, Portland, OR, USA,)(PODS’20). Association for Computing Machinery, New York, NY, USA, 1–16. doi:10.1145/3375395.3387641

    work page doi:10.1145/3375395.3387641 2020

  19. [19]

    Hao Guo and Youyou Lu. 2025. Achieving Low-Latency Graph-Based Vector Search via Aligning Best-First Search Algorithm with SSD. In 19th USENIX Symposium on Operating Systems Design and Implemen- tation (OSDI ’25). USENIX Association, Boston, MA, USA

    work page 2025

  20. [20]

    Yupeng Hou, Jiacheng Li, Zhankui He, An Yan, Xiusi Chen, and Ju- lian J. McAuley. 2024. Bridging Language and Items for Retrieval and Recommendation.CoRRabs/2403.03952 (2024). arXiv:2403.03952 doi:10.48550/ARXIV.2403.03952

    work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2403.03952 2024

  21. [21]

    Haodi Jiang, Hao Guo, Minhui Xie, Jiwu Shu, and Youyou Lu. 2026. High-Throughput, Cost-Effective Billion-Scale Vector Search with a Single GPU. InProceedings of the 2026 International Conference on Man- agement of Data (SIGMOD ’26). Association for Computing Machinery, Bengaluru, India

    work page 2026

  22. [22]

    Dingyi Kang, Dongming Jiang, Hanshen Yang, Hang Liu, and Bingzhe Li. 2025. Scalable Disk-Based Approximate Nearest Neighbor Search with Page-Aligned Graph. arXiv:2509.25487 [cs.LG]https://arxiv.org/ abs/2509.25487

    work page arXiv 2025

  23. [23]

    Patrick Lewis, Ethan Perez, Aleksandra Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Küttler, Mike Lewis, Wen-tau Yih, Tim Rocktäschel, Sebastian Riedel, and Douwe Kiela. 2020. Retrieval- augmented generation for knowledge-intensive NLP tasks. InPro- ceedings of the 34th International Conference on Neural Information Processing Systems ...

    work page 2020

  24. [24]

    Andersen, and Yuxiong He

    Conglong Li, Minjia Zhang, David G. Andersen, and Yuxiong He. 2020. Improving Approximate Nearest Neighbor Search through Learned Adaptive Early Termination. InProceedings of the 2020 ACM SIGMOD International Conference on Management of Data (SIGMOD ’20). As- sociation for Computing Machinery, Portland, OR, USA, 2539–2554. doi:10.1145/3318464.3380600 13

    work page doi:10.1145/3318464.3380600 2020

  25. [25]

    Jie Li, Haifeng Liu, Chuanghua Gui, Jianyu Chen, Zhenyuan Ni, Ning Wang, and Yuan Chen. 2018. The Design and Implementation of a Real Time Visual Search System on JD E-commerce Platform. InProceedings of the 19th International Middleware Conference Industry (Middleware ’18). Association for Computing Machinery, Rennes, France, 9–16. doi:10.1145/3284028.3284030

    work page doi:10.1145/3284028.3284030 2018

  26. [26]

    Mocheng Li, Xiao Yan, Baotong Lu, Yue Zhang, James Cheng, and Chenhao Ma. 2026. Attribute Filtering in Approximate Nearest Neigh- bor Search: An In-depth Experimental Study. InProceedings of the 2026 International Conference on Management of Data (SIGMOD ’26). Association for Computing Machinery, Bengaluru, India

    work page 2026

  27. [27]

    Anqi Liang, Pengcheng Zhang, Bin Yao, Zhongpu Chen, Yitong Song, and Guangxu Cheng. 2024. UNIFY: Unified Index for Range Filtered Approximate Nearest Neighbors Search.Proceedings of the VLDB Endowment(2024), 1118–1130. doi:10.14778/3717755.3717770

    work page doi:10.14778/3717755.3717770 2024

  28. [28]

    Malkov and D

    Yu A. Malkov and D. A. Yashunin. 2020. Efficient and Robust Approx- imate Nearest Neighbor Search Using Hierarchical Navigable Small World Graphs.IEEE Transactions on Pattern Analysis and Machine Intel- ligence (TPAMI)42, 4 (2020), 824–836. doi:10.1109/TPAMI.2018.2889473

    work page doi:10.1109/tpami.2018.2889473 2020

  29. [29]

    Tomás Mikolov, Kai Chen, Greg Corrado, and Jeffrey Dean. 2013. Ef- ficient Estimation of Word Representations in Vector Space. In1st International Conference on Learning Representations, Workshop Track Proceedings (ICLR ’13). Scottsdale, Arizona, USA.http://arxiv.org/abs/ 1301.3781

    work page internal anchor Pith review Pith/arXiv arXiv 2013

  30. [30]

    Milvus. 2026. IVF_PQ.https://milvus.io/docs/ivf-pq.md

    work page 2026

  31. [31]

    Liana Patel, Peter Kraft, Carlos Guestrin, and Matei Zaharia. 2024. ACORN: Performant and Predicate-Agnostic Search Over Vector Em- beddings and Structured Data. , Article 120 (2024). doi:10.1145/3654923

    work page doi:10.1145/3654923 2024

  32. [32]

    Alec Radford, Jong Wook Kim, Chris Hallacy, Aditya Ramesh, Gabriel Goh, Sandhini Agarwal, Girish Sastry, Amanda Askell, Pamela Mishkin, Jack Clark, Gretchen Krueger, and Ilya Sutskever. 2021. Learning Transferable Visual Models From Natural Language Supervision. In Proceedings of the 38th International Conference on Machine Learning (ICML ’21). PMLR, Virt...

    work page 2021

  33. [33]

    Christoph Schuhmann, Richard Vencu, Romain Beaumont, Robert Kaczmarczyk, Clayton Mullis, Aarush Katta, Theo Coombes, Jenia Jitsev, and Aran Komatsuzaki. 2021. LAION-400M: Open Dataset of CLIP-Filtered 400 Million Image-Text Pairs.CoRRabs/2111.02114 (2021). arXiv:2111.02114https://arxiv.org/abs/2111.02114

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  34. [34]

    Griffiths Selinger, M

    P. Griffiths Selinger, M. M. Astrahan, D. D. Chamberlin, R. A. Lo- rie, and T. G. Price. 1979. Access path selection in a relational data- base management system. InProceedings of the 1979 ACM SIGMOD International Conference on Management of Data (SIGMOD ’79). As- sociation for Computing Machinery, Boston, Massachusetts, 23–34. doi:10.1145/582095.582099

    work page doi:10.1145/582095.582099 1979

  35. [35]

    Harsha Simhadri. 2021. Results of the NeurIPS’21 Challenge on Billion- Scale Approximate Nearest Neighbor Search. InProceedings of the 35th International Conference on Neural Information Processing Systems (NIPS ’21). Curran Associates Inc., Red Hook, NY, USA

    work page 2021

  36. [36]

    Harsha Simhadri. 2022. Research talk: Approximate nearest neighbor search systems at scale.https://www.youtube.com/watch?v=BnYNdS IKibQ&list=PLD7HFcN7LXReJTWFKYqwMcCc1nZKIXBo9&index= 9

    work page 2022

  37. [37]

    Harsha Vardhan simhadri, Martin Aumüller, Matthijs Douze, Dmitry Baranchuk, Amir Ingber, Edo Liberty, George Williams, Ben Landrum, Magdalen Dobson Manohar, Mazin Karjikar, Laxman Dhulipala, Meng Chen, Yue Chen, Rui Ma, Kai Zhang, Yuzheng Cai, Jiayang Shi, Weiguo Zheng, Yizhuo Chen, Jie Yin, and Ben Huang. 2025. Results of the Big ANN: NeurIPS’23 competit...

    work page 2025

  38. [38]

    Haoyu Song, Sarang Dharmapurikar, Jonathan Turner, and John Lock- wood. 2005. Fast hash table lookup using extended bloom filter: an aid to network processing. InProceedings of the 2005 Conference on Applications, Technologies, Architectures, and Protocols for Computer Communications (SIGCOMM ’05). Association for Computing Machin- ery, Philadelphia, Penn...

    work page doi:10.1145/1080091 2005

  39. [39]

    Suhas Jayaram Subramanya, Devvrit, Rohan Kadekodi, Ravishankar Krishaswamy, and Harsha Vardhan Simhadri. 2019. DiskANN: fast accurate billion-point nearest neighbor search on a single node. In Proceedings of the 33rd International Conference on Neural Information Processing Systems (NIPS ’19). Curran Associates Inc., Red Hook, NY, USA, Article 1233

    work page 2019

  40. [40]

    Bing Tian, Haikun Liu, Zhuohui Duan, Xiaofei Liao, Hai Jin, and Yu Zhang. 2024. Scalable Billion-point Approximate Nearest Neighbor Search Using SmartSSDs. In2024 USENIX Annual Technical Conference (USENIX ATC ’24). USENIX Association, Santa Clara, CA, 1135–1150. https://www.usenix.org/conference/atc24/presentation/tian

    work page 2024

  41. [41]

    Bing Tian, Haikun Liu, Yuhang Tang, Shihai Xiao, Zhuohui Duan, Xiaofei Liao, Hai Jin, Xuecang Zhang, Junhua Zhu, and Yu Zhang

    work page

  42. [42]

    In23rd USENIX Conference on File and Storage Technologies (FAST ’25)

    Towards High-throughput and Low-latency Billion-scale Vector Search via CPU/GPU Collaborative Filtering and Re-ranking. In23rd USENIX Conference on File and Storage Technologies (FAST ’25). USENIX Association, Santa Clara, CA, 171–185.https://www.usenix.org/con ference/fast25/presentation/tian-bing

    work page

  43. [43]

    Toussaint

    Godfried T. Toussaint. 1980. The relative neighbourhood graph of a finite planar set.Pattern Recognition12, 4 (1980), 261–268. doi:10.101 6/0031-3203(80)90066-7

    work page 1980

  44. [44]

    Jianguo Wang, Xiaomeng Yi, Rentong Guo, Hai Jin, Peng Xu, Shengjun Li, Xiangyu 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. InProceedings of the 2021 ...

    work page doi:10.1145/3448016.3457550 2021

  45. [45]

    Mengzhao Wang, Lingwei Lv, Xiaoliang Xu, Yuxiang Wang, Qiang Yue, and Jiongkang Ni. 2023. An Efficient and Robust Framework for Approximate Nearest Neighbor Search with Attribute Constraint. In Advances in Neural Information Processing Systems, A. Oh, T. Naumann, A. Globerson, K. Saenko, M. Hardt, and S. Levine (Eds.), Vol. 36. Curran Associates, Inc., 15...

    work page 2023

  46. [46]

    Mengzhao Wang, Weizhi Xu, Xiaomeng Yi, Songlin Wu, Zhangyang Peng, Xiangyu Ke, Yunjun Gao, Xiaoliang Xu, Rentong Guo, and Charles Xie. 2024. Starling: An I/O-Efficient Disk-Resident Graph Index Framework for High-Dimensional Vector Similarity Search on Data Segment. InProceedings of the ACM on Management of Data (SIGMOD ’24). Association for Computing Mac...

    work page doi:10.1145/3639269 2024

  47. [47]

    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. In Proceedings of the VLDB Endowment (VLDB ’20). VLDB Endowment, Tokyo, Japan, 3152–3165. doi:10.14778/3415478.3415541

    work page doi:10.14778/3415478.3415541 2020

  48. [48]

    Yuexuan Xu, Jianyang Gao, Yutong Gou, Cheng Long, and Christian S. Jensen. 2024. iRangeGraph: Improvising Range-dedicated Graphs for Range-filtering Nearest Neighbor Search. InProceedings of the ACM on Management of Data (SIGMOD ’24). Association for Computing Machinery, Santiago, Chile. doi:10.1145/3698814

    work page doi:10.1145/3698814 2024

  49. [49]

    Yuming Xu, Hengyu Liang, Jin Li, Shuotao Xu, Qi Chen, Qianxi Zhang, Cheng Li, Ziyue Yang, Fan Yang, Yuqing Yang, Peng Cheng, and Mao Yang. 2023. SPFresh: Incremental In-Place Update for Billion-Scale Vector Search. InProceedings of the 29th Symposium on Operating Systems Principles (SOSP ’23). Association for Computing Machinery, 14 Koblenz, Germany, 545–...

    work page doi:10.1145/3600006.3613166 2023

  50. [50]

    Peiqi Yin, Xiao Yan, Qihui Zhou, Hui Li, Xiaolu Li, Lin Zhang, Meiling Wang, Xin Yao, and James Cheng. 2025. Gorgeous: Revisiting the Data Layout for Disk-Resident High-Dimensional Vector Search.arXiv preprint arXiv:2508.15290(2025)

    work page arXiv 2025

  51. [51]

    Minlan Yu, Alex Fabrikant, and Jennifer Rexford. 2009. BUFFALO: bloom filter forwarding architecture for large organizations. InPro- ceedings of the 5th International Conference on Emerging Networking Ex- periments and Technologies (CoNEXT ’09). Association for Computing Machinery, New York, NY, USA, 313–324. doi:10.1145/1658939.1658975

    work page doi:10.1145/1658939.1658975 2009

  52. [52]

    Andersen, Michael Kaminsky, Kimberly Keeton, and Andrew Pavlo

    Huanchen Zhang, Hyeontaek Lim, Viktor Leis, David G. Andersen, Michael Kaminsky, Kimberly Keeton, and Andrew Pavlo. 2018. SuRF: Practical Range Query Filtering with Fast Succinct Tries. InProceed- ings of the 2018 International Conference on Management of Data (SIG- MOD ’18). Association for Computing Machinery, Houston, TX, USA, 323–336. doi:10.1145/3183...

    work page doi:10.1145/3183713.3196931 2018

  53. [53]

    Chaoji Zuo, Miao Qiao, Wenchao Zhou, Feifei Li, and Dong Deng

    work page

  54. [54]

    InProceedings of the ACM on Management of Data (SIGMOD ’24)

    SeRF: Segment Graph for Range-Filtering Approximate Nearest Neighbor Search. InProceedings of the ACM on Management of Data (SIGMOD ’24). Association for Computing Machinery, Santiago, Chile, Article 69. doi:10.1145/3639324 15

    work page doi:10.1145/3639324