pith. sign in

arxiv: 2503.13822 · v3 · submitted 2025-03-18 · 💻 cs.DB

NeurBench: A Benchmark Suite for Learned Database Components with Drift Modeling

Zhanhao Zhao , Haotian Gao , Naili Xing , Lingze Zeng , Meihui Zhang , Gang Chen , Manuel Rigger , Beng Chin Ooi This is my paper

Pith reviewed 2026-05-23 00:06 UTC · model grok-4.3

classification 💻 cs.DB
keywords learned database componentsdata driftworkload driftbenchmark suitedrift factordrift-aware generationrobustness evaluationNeurBench
0
0 comments X

The pith

NeurBench introduces a drift factor and generation framework to evaluate learned database components under measurable and controllable data and workload drift.

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

The paper presents NeurBench as a benchmark suite for testing learned database components when data and workloads change over time. It defines a drift factor to quantify different kinds of drift in a measurable way. A drift-aware framework then generates data and workloads that aim to mimic real changes while keeping natural correlations intact. Experiments confirm the generated drifts look realistic and reveal how representative components behave across drift scenarios. This setup allows more systematic checks of robustness than previous targeted tests.

Core claim

NeurBench is a benchmark suite that quantifies diverse types of drift via a drift factor and supplies a drift-aware data and workload generation framework that simulates real-world drift while preserving inherent correlations, thereby enabling systematic performance evaluation of learned database components under a broad range of measurable and controllable drift conditions.

What carries the argument

The drift factor, which quantifies diverse types of drift, together with the drift-aware data and workload generation framework built upon it.

If this is right

  • Learned components can be tested across a broad range of drift scenarios rather than only specific cases.
  • Customized drift conditions become available for targeted evaluation.
  • Performance insights under varying drift become reproducible and comparable.
  • Robustness assessment of learned components moves from ad-hoc checks to systematic coverage.

Where Pith is reading between the lines

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

  • The same drift-factor approach could be adapted to benchmark learned components in other dynamic systems such as caching or query optimizers.
  • Standard test suites for learned systems might incorporate drift factors as a default dimension.
  • Component designers could use the generation framework to create training data that anticipates future drift.
  • Comparative studies across multiple learned components would become feasible under identical drift conditions.

Load-bearing premise

The drift-aware data and workload generation framework effectively simulates real-world drift while preserving inherent correlations.

What would settle it

A direct statistical comparison in which the correlations and distributional properties of data and workloads generated by the framework diverge from those observed in actual production database traces under documented drift.

Figures

Figures reproduced from arXiv: 2503.13822 by Beng Chin Ooi, Gang Chen, Haotian Gao, Lingze Zeng, Manuel Rigger, Meihui Zhang, Naili Xing, Zhanhao Zhao.

Figure 1
Figure 1. Figure 1: System Overview of NeurBench factor of 𝑑=0 indicates no drift, while 𝑑=1 denotes that the entire data or workload is fully drifted. Section 4.1 provides the formal definition of the drift factor. Given a specified drift factor 𝑑 as input, the idea is to drift the given original data (workload) to a target data (workload) that aligns with 𝑑, thereby enabling controllable drifted data and workload generation… view at source ↗
Figure 2
Figure 2. Figure 2: Diffuser and Drifter Training guides the process to introduce the specified drift 𝑑. Since the drifted data x𝑑𝑟𝑖 𝑓 𝑡 is the target output, Equation (2) can be rewritten as: 𝑝(x𝑡−1 |x𝑡 , x𝑑𝑟𝑖 𝑓 𝑡) ∝ 𝑝(x𝑡−1 |x𝑡)Pr(x𝑑𝑟𝑖 𝑓 𝑡 |x𝑡−1, x𝑡). (3) Note that the mean of the Gaussian distribution, 𝝁(x𝑡 , 𝑡; 𝜃), deter￾mines the trajectory of the denoising process [11, 36]. We therefore leverage Pr(x𝑑𝑟𝑖 𝑓 𝑡 |x𝑡−1, x𝑡) to… view at source ↗
Figure 3
Figure 3. Figure 3: End-to-end Learned Query Optimizers Tree-LSTM [49]. Both structures are deep neural networks designed for tree structures to capture bottom-up sequential information. Plan Search. There are two paradigms in plan search, which differ in whether they rely on traditional query optimizers: 1) Black-box search provides auxiliary information to traditional query optimizers for candidate plan generation. Based on… view at source ↗
Figure 4
Figure 4. Figure 4: Structures of Existing Learned Indexes the design of the index structure and update mechanism are key factors influencing the performance of learned indexes. Index Structure. As shown in [PITH_FULL_IMAGE:figures/full_fig_p008_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Performance on Generating Drifted Datasets with Varying Drift Factors [PITH_FULL_IMAGE:figures/full_fig_p009_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Performance on Generating Drifted Workloads [PITH_FULL_IMAGE:figures/full_fig_p010_6.png] view at source ↗
Figure 8
Figure 8. Figure 8: Performance Evaluation of Learned Query Optimizers using Drifted CH-Benchmark Workload from adapting to unseen join patterns and leading to inaccurate latency predictions. Neo, which employs a white-box plan search with a best-only plan selection strategy, is more susceptible to local optima and thus struggles to explore globally optimal plans under join pattern drift. In contrast, as discussed in O2, Lero… view at source ↗
Figure 7
Figure 7. Figure 7: Performance of Learned Query Optimizers under [PITH_FULL_IMAGE:figures/full_fig_p011_7.png] view at source ↗
Figure 9
Figure 9. Figure 9: Performance of Learned Indexes under Data Drift [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Performance of Learned Indexes with Concurrent [PITH_FULL_IMAGE:figures/full_fig_p013_10.png] view at source ↗
Figure 12
Figure 12. Figure 12: Throughput of Learned CC under Arrival Rate [PITH_FULL_IMAGE:figures/full_fig_p013_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: CCaaLF vs PolyJuice (Arrival Rate Drift with Drift [PITH_FULL_IMAGE:figures/full_fig_p014_13.png] view at source ↗
read the original abstract

Learned database components, which deeply integrate machine learning into their design, have been extensively studied in recent years. Given the dynamism of databases, where data and workloads continuously drift, it is crucial for learned database components to remain effective and efficient in the face of data and workload drift. Robustness, therefore, is a key factor in assessing their practical applicability. Although recent works examine learned database components under specific drift, they fail to enable systematic performance evaluations across a broad range of drift or under customized drift as needed. This paper presents NeurBench, a new benchmark suite that supports evaluating learned database components under measurable and controllable data and workload drift. We quantify diverse types of drift by introducing a key concept called the drift factor. Building on this formulation, we propose a drift-aware data and workload generation framework that effectively simulates real-world drift while preserving inherent correlations. Experimental results demonstrate the effectiveness of NeurBench in generating realistic data and workload drift, while providing insights into the performance of representative learned database components under different drift scenarios.

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

0 major / 2 minor

Summary. The paper introduces NeurBench, a benchmark suite for evaluating learned database components under data and workload drift. It defines a drift factor to quantify diverse drift types and proposes a drift-aware data and workload generation framework claimed to simulate real-world drift while preserving inherent correlations. Experimental results are presented to demonstrate the framework's effectiveness in generating realistic drift scenarios and to provide insights into the performance of representative learned database components under varying drift conditions.

Significance. If the experimental validation holds, NeurBench would fill a notable gap by enabling systematic, measurable, and customizable evaluation of robustness to drift in learned database systems, moving beyond ad-hoc or single-drift studies. The drift factor and generation framework, if shown to preserve correlations, could support reproducible comparisons and development of drift-resilient components, with the reported insights adding immediate practical value to the field.

minor comments (2)
  1. [Abstract] Abstract: the phrase 'diverse types of drift' is used without enumeration or examples; a short parenthetical list of the quantified drift types would improve immediate readability.
  2. [Abstract] Abstract: the claim of 'preserving inherent correlations' is central to the framework's value; ensure the experiments section includes explicit quantitative metrics (e.g., correlation coefficients before/after generation) rather than qualitative statements alone.

Simulated Author's Rebuttal

0 responses · 0 unresolved

We thank the referee for the positive summary, significance assessment, and recommendation of minor revision. No major comments were raised in the report.

Circularity Check

0 steps flagged

No significant circularity

full rationale

The paper introduces NeurBench as a benchmark suite with a drift factor concept and a drift-aware generation framework. No derivation chain, predictions, or first-principles results are claimed that reduce to fitted parameters or self-citations by construction. The central contribution is the design and experimental validation of the benchmark tool itself, which is self-contained and externally falsifiable via its generated data/workloads. No load-bearing self-citations, ansatzes, or renamings of known results appear in the provided text.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 1 invented entities

The central claim depends on the new drift factor being a valid quantification and the generation framework producing realistic drift that preserves correlations; these are introduced without external validation in the abstract.

axioms (1)
  • domain assumption The proposed drift-aware generation framework preserves inherent correlations in data and workloads when simulating drift.
    Invoked to ensure simulated drift is realistic for evaluation purposes.
invented entities (1)
  • drift factor no independent evidence
    purpose: To quantify diverse types of drift in a measurable and controllable manner.
    New concept introduced to enable the benchmark's systematic evaluation capability.

pith-pipeline@v0.9.0 · 5725 in / 1195 out tokens · 47372 ms · 2026-05-23T00:06:40.438306+00:00 · methodology

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

74 extracted references · 74 canonical work pages

  1. [1]

    Cohen, Zhenggang Xu, Prithviraj Pandian, Nikolay Laptev, and Ryan Marcus

    Christoph Anneser, Nesime Tatbul, David E. Cohen, Zhenggang Xu, Prithviraj Pandian, Nikolay Laptev, and Ryan Marcus. 2023. AutoSteer: Learned Query Optimization for Any SQL Database.Proc. VLDB Endow. 16, 12 (2023), 3515–3527

    work page 2023

  2. [2]

    Internet Archive. 2025. IMDB Archive . https://web.archive.org/web/ 20240301000000*/https://datasets.imdbws.com/

    work page 2025

  3. [3]

    Lawrence Benson, Carsten Binnig, Jan-Micha Bodensohn, Federico Lorenzi, Jigao Luo, Danica Porobic, Tilmann Rabl, Anupam Sanghi, Russell Sears, Pinar Tözün, and Tobias Ziegler. 2024. Surprise Benchmarking: The Why, What, and How. In DBTest@SIGMOD. ACM, 1–8

    work page 2024

  4. [4]

    Maximilian Böther, Foteini Strati, Viktor Gsteiger, and Ana Klimovic. 2023. To- wards A Platform and Benchmark Suite for Model Training on Dynamic Datasets. In EuroMLSys@EuroSys. ACM, 8–17

    work page 2023

  5. [5]

    Tianyi Chen, Jun Gao, Hedui Chen, and Yaofeng Tu. 2023. LOGER: A Learned Optimizer towards Generating Efficient and Robust Query Execution Plans. Proc. VLDB Endow. 16, 7 (2023), 1777–1789

    work page 2023

  6. [6]

    Yannis Chronis, Yawen Wang, Yu Gan, Sami Abu-El-Haija, Chelsea Lin, Carsten Binnig, and Fatma Özcan. 2024. CardBench: A Benchmark for Learned Cardinality Estimation in Relational Databases. CoRR abs/2408.16170 (2024)

    work page arXiv 2024

  7. [7]

    Richard Cole, Florian Funke, Leo Giakoumakis, Wey Guy, Alfons Kemper, Stefan Krompass, Harumi Kuno, Raghunath Nambiar, Thomas Neumann, Meikel Poess, et al. 2011. The mixed workload CH-benCHmark. In Proceedings of the Fourth International Workshop on Testing Database Systems. 1–6

    work page 2011

  8. [8]

    Cooper, Adam Silberstein, Erwin Tam, Raghu Ramakrishnan, and Russell Sears

    Brian F. Cooper, Adam Silberstein, Erwin Tam, Raghu Ramakrishnan, and Russell Sears. 2010. Benchmarking cloud serving systems with YCSB. In SoCC. ACM, 143–154

    work page 2010

  9. [9]

    The Transaction Processing Council. 2024. TPC-C. http://www.tpc.org/tpcc/

    work page 2024

  10. [10]

    The Transaction Processing Council. 2024. TPC-H. http://www.tpc.org/tpch/

    work page 2024

  11. [11]

    Prafulla Dhariwal and Alexander Quinn Nichol. 2021. Diffusion Models Beat GANs on Image Synthesis. In NeurIPS. 8780–8794

    work page 2021

  12. [12]

    Narasayya

    Bailu Ding, Surajit Chaudhuri, Johannes Gehrke, and Vivek R. Narasayya. 2021. DSB: A Decision Support Benchmark for Workload-Driven and Traditional Database Systems. Proc. VLDB Endow. 14, 13 (2021), 3376–3388

    work page 2021

  13. [13]

    Lomet, and Tim Kraska

    Jialin Ding, Umar Farooq Minhas, Jia Yu, Chi Wang, Jaeyoung Do, Yinan Li, Hantian Zhang, Badrish Chandramouli, Johannes Gehrke, Donald Kossmann, David B. Lomet, and Tim Kraska. 2020. ALEX: An Updatable Adaptive Learned Index. In SIGMOD Conference. ACM, 969–984

    work page 2020

  14. [14]

    Paolo Ferragina and Giorgio Vinciguerra. 2020. The PGM-index: a fully-dynamic compressed learned index with provable worst-case bounds. Proc. VLDB Endow. 13, 8 (2020), 1162–1175

    work page 2020

  15. [15]

    Josh Gardner, Zoran Popovic, and Ludwig Schmidt. 2023. Benchmarking Distri- bution Shift in Tabular Data with TableShift. In NeurIPS

    work page 2023

  16. [16]

    Butts, and Athina Markopoulou

    Minas Gjoka, Maciej Kurant, Carter T. Butts, and Athina Markopoulou. 2010. Walking in Facebook: A Case Study of Unbiased Sampling of OSNs. InINFOCOM. IEEE, 2498–2506

    work page 2010

  17. [17]

    Zhihan Guo, Kan Wu, Cong Yan, and Xiangyao Yu. 2021. Releasing Locks As Early As You Can: Reducing Contention of Hotspots by Violating Two-Phase Locking. In SIGMOD Conference. ACM, 658–670

    work page 2021

  18. [18]

    Yuxing Han, Ziniu Wu, Peizhi Wu, Rong Zhu, Jingyi Yang, Liang Wei Tan, Kai Zeng, Gao Cong, Yanzhao Qin, Andreas Pfadler, Zhengping Qian, Jingren Zhou, Jiangneng Li, and Bin Cui. 2021. Cardinality Estimation in DBMS: A Comprehensive Benchmark Evaluation. Proc. VLDB Endow. 15, 4 (2021), 752– 765

    work page 2021

  19. [19]

    Jonathan Ho, Ajay Jain, and Pieter Abbeel. 2020. Denoising Diffusion Probabilistic Models. In NeurIPS

    work page 2020

  20. [20]

    IMDB. 2025. IMDB Dataset. https://www.imdb.com/

    work page 2025

  21. [21]

    Haque, Sara M

    Pang Wei Koh, Shiori Sagawa, Henrik Marklund, Sang Michael Xie, Marvin Zhang, Akshay Balsubramani, Weihua Hu, Michihiro Yasunaga, Richard Lanas Phillips, Irena Gao, Tony Lee, Etienne David, Ian Stavness, Wei Guo, Berton Earn- shaw, Imran S. Haque, Sara M. Beery, Jure Leskovec, Anshul Kundaje, Emma Pierson, Sergey Levine, Chelsea Finn, and Percy Liang. 202...

    work page 2021

  22. [22]

    Akim Kotelnikov, Dmitry Baranchuk, Ivan Rubachev, and Artem Babenko. 2023. TabDDPM: Modelling Tabular Data with Diffusion Models. In ICML (Proceedings of Machine Learning Research, Vol. 202) . PMLR, 17564–17579

    work page 2023

  23. [23]

    Chi, Jeffrey Dean, and Neoklis Polyzotis

    Tim Kraska, Alex Beutel, Ed H. Chi, Jeffrey Dean, and Neoklis Polyzotis. 2018. The Case for Learned Index Structures. In SIGMOD Conference. ACM, 489–504

    work page 2018

  24. [24]

    Meghdad Kurmanji, Eleni Triantafillou, and Peter Triantafillou. 2024. Machine Unlearning in Learned Databases: An Experimental Analysis. Proc. ACM Manag. Data 2, 1 (2024), 49:1–49:26

    work page 2024

  25. [25]

    Meghdad Kurmanji and Peter Triantafillou. 2023. Detect, Distill and Update: Learned DB Systems Facing Out of Distribution Data. Proc. ACM Manag. Data 1, 1 (2023), 33:1–33:27

    work page 2023

  26. [26]

    Shane Culpepper, and Renata Borovica-Gajic

    Hai Lan, Zhifeng Bao, J. Shane Culpepper, and Renata Borovica-Gajic. 2023. Updatable Learned Indexes Meet Disk-Resident DBMS - From Evaluations to Design Choices. Proc. ACM Manag. Data 1, 2 (2023), 139:1–139:22

    work page 2023

  27. [27]

    Claude Lehmann, Pavel Sulimov, and Kurt Stockinger. 2024. Is Your Learned Query Optimizer Behaving As You Expect? A Machine Learning Perspective. Proc. VLDB Endow. 17, 7 (2024), 1565–1577. NeurBench: Benchmarking Learned Database Components with Data and Workload Drift Modeling

    work page 2024

  28. [28]

    Boncz, Alfons Kemper, and Thomas Neumann

    Viktor Leis, Andrey Gubichev, Atanas Mirchev, Peter A. Boncz, Alfons Kemper, and Thomas Neumann. 2015. How Good Are Query Optimizers, Really? Proc. VLDB Endow. 9, 3 (2015), 204–215

    work page 2015

  29. [29]

    Viktor Leis, Alfons Kemper, and Thomas Neumann. 2013. The adaptive radix tree: ARTful indexing for main-memory databases. In ICDE. IEEE Computer Society, 38–49

    work page 2013

  30. [30]

    Viktor Leis, Florian Scheibner, Alfons Kemper, and Thomas Neumann. 2016. The ART of practical synchronization. In DaMoN. ACM, 3:1–3:8

    work page 2016

  31. [31]

    Beibin Li, Yao Lu, and Srikanth Kandula. 2022. Warper: Efficiently Adapting Learned Cardinality Estimators to Data and Workload Drifts. In SIGMOD Con- ference. ACM, 1920–1933

    work page 2022

  32. [32]

    Pengfei Li, Wenqing Wei, Rong Zhu, Bolin Ding, Jingren Zhou, and Hua Lu. 2023. ALECE: An Attention-based Learned Cardinality Estimator for SPJ Queries on Dynamic Workloads. Proc. VLDB Endow. 17, 2 (2023), 197–210

    work page 2023

  33. [33]

    Wendi Li, Xiao Yang, Weiqing Liu, Yingce Xia, and Jiang Bian. 2022. DDG-DA: Data Distribution Generation for Predictable Concept Drift Adaptation. In AAAI. AAAI Press, 4092–4100

    work page 2022

  34. [34]

    Yu-Shan Lin, Ching Tsai, Tz-Yu Lin, Yun-Sheng Chang, and Shan-Hung Wu

    work page

  35. [35]

    In SIGMOD Conference

    Don’t Look Back, Look into the Future: Prescient Data Partitioning and Migration for Deterministic Database Systems. In SIGMOD Conference. ACM, 1156–1168

    work page

  36. [36]

    Mandic, Wenwu Wang, and Mark D

    Haohe Liu, Zehua Chen, Yi Yuan, Xinhao Mei, Xubo Liu, Danilo P. Mandic, Wenwu Wang, and Mark D. Plumbley. 2023. AudioLDM: Text-to-Audio Gener- ation with Latent Diffusion Models. In ICML (Proceedings of Machine Learning Research, Vol. 202). PMLR, 21450–21474

    work page 2023

  37. [37]

    Tongyu Liu, Ju Fan, Nan Tang, Guoliang Li, and Xiaoyong Du. 2024. Controllable Tabular Data Synthesis Using Diffusion Models. Proc. ACM Manag. Data 2, 1 (2024), 28:1–28:29

    work page 2024

  38. [38]

    Chaohong Ma, Xiaohui Yu, Yifan Li, Xiaofeng Meng, and Aishan Maoliniyazi

    work page

  39. [39]

    FILM: a Fully Learned Index for Larger-than-Memory Databases. Proc. VLDB Endow. 16, 3 (2022), 561–573

    work page 2022

  40. [40]

    Manning and Hinrich Schütze

    Christopher D. Manning and Hinrich Schütze. 2001. Foundations of statistical natural language processing. MIT Press

    work page 2001

  41. [41]

    Ryan Marcus, Parimarjan Negi, Hongzi Mao, Nesime Tatbul, Mohammad Al- izadeh, and Tim Kraska. 2021. Bao: Making Learned Query Optimization Practical. In SIGMOD Conference. ACM, 1275–1288

    work page 2021

  42. [42]

    Ryan Marcus, Parimarjan Negi, Hongzi Mao, Chi Zhang, Mohammad Alizadeh, Tim Kraska, Olga Papaemmanouil, and Nesime Tatbul. 2019. Neo: A Learned Query Optimizer. Proc. VLDB Endow. 12, 11 (2019), 1705–1718

    work page 2019

  43. [43]

    Songsong Mo, Yile Chen, Hao Wang, Gao Cong, and Zhifeng Bao. 2023. Lemo: A Cache-Enhanced Learned Optimizer for Concurrent Queries. Proc. ACM Manag. Data 1, 4 (2023), 247:1–247:26

    work page 2023

  44. [44]

    Lili Mou, Ge Li, Lu Zhang, Tao Wang, and Zhi Jin. 2016. Convolutional Neural Networks over Tree Structures for Programming Language Processing. In AAAI. AAAI Press, 1287–1293

    work page 2016

  45. [45]

    NeurBench. 2025. NeurBench Implementation . https://github.com/neurdb/ neurbench

    work page 2025

  46. [46]

    Hexiang Pan, Shaofeng Cai, Yeow Meng Chee, Tien Tuan Anh Dinh, Yuncheng Wu, and Beng Chin Ooi. 2025. CCaaLF: Concurrency Control as a Learnable Function. CoRR abs/2503.10036 (2025)

    work page arXiv 2025

  47. [47]

    Piotr Porwik and Benjamin Mensah Dadzie. 2022. Detection of data drift in a two-dimensional stream using the Kolmogorov-Smirnov test. In KES (Procedia Computer Science, Vol. 207). Elsevier, 168–175

    work page 2022

  48. [48]

    Cemal Okan Sakar, Suleyman Olcay Polat, Mete Katircioglu, and Yomi Kastro

    work page

  49. [49]

    Neural Comput

    Real-time prediction of online shoppers’ purchasing intention using multi- layer perceptron and LSTM recurrent neural networks. Neural Comput. Appl. 31, 10 (2019), 6893–6908

    work page 2019

  50. [50]

    Weiss, Niru Maheswaranathan, and Surya Ganguli

    Jascha Sohl-Dickstein, Eric A. Weiss, Niru Maheswaranathan, and Surya Ganguli

    work page

  51. [51]

    In Proceedings of the 32nd International Conference on Machine Learning, ICML 2015, Lille, France, 6-11 July 2015 (JMLR Workshop and Conference Proceedings, Vol

    Deep Unsupervised Learning using Nonequilibrium Thermodynamics. In Proceedings of the 32nd International Conference on Machine Learning, ICML 2015, Lille, France, 6-11 July 2015 (JMLR Workshop and Conference Proceedings, Vol. 37) , Francis R. Bach and David M. Blei (Eds.). JMLR.org, 2256–2265

    work page 2015

  52. [52]

    Zhaoyan Sun, Xuanhe Zhou, and Guoliang Li. 2023. Learned Index: A Compre- hensive Experimental Evaluation. Proc. VLDB Endow. 16, 8 (2023), 1992–2004

    work page 2023

  53. [53]

    Kai Sheng Tai, Richard Socher, and Christopher D. Manning. 2015. Improved Semantic Representations From Tree-Structured Long Short-Term Memory Net- works. In ACL (1). The Association for Computer Linguistics, 1556–1566

    work page 2015

  54. [54]

    Chuzhe Tang, Youyun Wang, Zhiyuan Dong, Gansen Hu, Zhaoguo Wang, Minjie Wang, and Haibo Chen. 2020. XIndex: a scalable learned index for multicore data storage. In PPoPP. ACM, 308–320

    work page 2020

  55. [55]

    Dixin Tang, Hao Jiang, and Aaron J. Elmore. 2017. Adaptive Concurrency Control: Despite the Looking Glass, One Concurrency Control Does Not Fit All. In CIDR. www.cidrdb.org

    work page 2017

  56. [56]

    Gomez, Lukasz Kaiser, and Illia Polosukhin

    Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N. Gomez, Lukasz Kaiser, and Illia Polosukhin. 2017. Attention is All you Need. In NIPS. 5998–6008

    work page 2017

  57. [57]

    Jia-Chen Wang, Ding Ding, Huan Wang, Conrad Christensen, Zhaoguo Wang, Haibo Chen, and Jinyang Li. 2021. Polyjuice: High-Performance Transactions via Learned Concurrency Control. In OSDI. USENIX Association, 198–216

    work page 2021

  58. [58]

    Zhaoguo Wang, Shuai Mu, Yang Cui, Han Yi, Haibo Chen, and Jinyang Li. 2016. Scaling Multicore Databases via Constrained Parallel Execution. In SIGMOD Conference. ACM, 1643–1658

    work page 2016

  59. [59]

    Andersen

    Ziqi Wang, Andrew Pavlo, Hyeontaek Lim, Viktor Leis, Huanchen Zhang, Michael Kaminsky, and David G. Andersen. 2018. Building a Bw-Tree Takes More Than Just Buzz Words. In SIGMOD Conference. ACM, 473–488

    work page 2018

  60. [60]

    WeatherAUS. 2025. WeatherAUS DataSet. https://www.kaggle.com/jsphyg/ weather-dataset-rattle-package

    work page 2025

  61. [61]

    Wikipedia. 2025. Pearson Correlation Coefficient. https://en.wikipedia.org/wiki/ Pearson_correlation_coefficient

    work page 2025

  62. [62]

    Olivia Wiles, Sven Gowal, Florian Stimberg, Sylvestre-Alvise Rebuffi, Ira Ktena, Krishnamurthy Dvijotham, and Ali Taylan Cemgil. 2022. A Fine-Grained Analysis on Distribution Shift. In ICLR. OpenReview.net

    work page 2022

  63. [63]

    Chaichon Wongkham, Baotong Lu, Chris Liu, Zhicong Zhong, Eric Lo, and Tianzheng Wang. 2022. Are Updatable Learned Indexes Ready? Proc. VLDB Endow. 15, 11 (2022), 3004–3017

    work page 2022

  64. [64]

    Jiacheng Wu, Yong Zhang, Shimin Chen, Yu Chen, Jin Wang, and Chunxiao Xing

    work page

  65. [65]

    Updatable Learned Index with Precise Positions. Proc. VLDB Endow. 14, 8 (2021), 1276–1288

    work page 2021

  66. [66]

    Peizhi Wu and Zachary G. Ives. 2024. Modeling Shifting Workloads for Learned Database Systems. Proc. ACM Manag. Data 2, 1 (2024), 38:1–38:27

    work page 2024

  67. [67]

    Yu Xia, Xiangyao Yu, Matthew Butrovich, Andrew Pavlo, and Srinivas Devadas

    work page

  68. [68]

    In SIGMOD Conference

    Litmus: Towards a Practical Database Management System with Verifiable ACID Properties and Transaction Correctness. In SIGMOD Conference. ACM, 1478–1492

    work page

  69. [69]

    Xiang Yu, Chengliang Chai, Guoliang Li, and Jiabin Liu. 2022. Cost-based or Learning-based? A Hybrid Query Optimizer for Query Plan Selection. Proc. VLDB Endow. 15, 13 (2022), 3924–3936

    work page 2022

  70. [70]

    Shunkang Zhang, Ji Qi, Xin Yao, and André Brinkmann. 2024. Hyper: A High- Performance and Memory-Efficient Learned Index via Hybrid Construction.Proc. ACM Manag. Data 2, 3 (2024), 145

    work page 2024

  71. [71]

    Zhou Zhang, Zhaole Chu, Peiquan Jin, Yongping Luo, Xike Xie, Shouhong Wan, Yun Luo, Xufei Wu, Peng Zou, Chunyang Zheng, Guoan Wu, and Andy Rudoff

    work page

  72. [72]

    PLIN: A Persistent Learned Index for Non-Volatile Memory with High Performance and Instant Recovery. Proc. VLDB Endow. 16, 2 (2022), 243–255

    work page 2022

  73. [73]

    Rong Zhu, Wei Chen, Bolin Ding, Xingguang Chen, Andreas Pfadler, Ziniu Wu, and Jingren Zhou. 2023. Lero: A Learning-to-Rank Query Optimizer. Proc. VLDB Endow. 16, 6 (2023), 1466–1479

    work page 2023

  74. [74]

    Rong Zhu, Lianggui Weng, Bolin Ding, and Jingren Zhou. 2024. Learned Query Optimizer: What is New and What is Next. In SIGMOD Conference Companion. ACM, 561–569

    work page 2024