pith. sign in

arxiv: 2605.25814 · v1 · pith:KTPI5SLZnew · submitted 2026-05-25 · 💻 cs.CL · cs.AI

Adaptive Graph Refinement and Label Propagation with LLMs for Cost-Effective Entity Resolution

Hongtao Wang , Renchi Yang , Haoran Zheng , Xiangyu Ke This is my paper

Pith reviewed 2026-06-29 21:27 UTC · model grok-4.3

classification 💻 cs.CL cs.AI
keywords entity resolutionlabel propagationgraph refinementlarge language modelsclusteringcost optimizationdirty dataprobabilistic methods
0
0 comments X

The pith

Alper unifies matching and clustering in entity resolution through iterative label propagation on an evolving graph that selectively incorporates LLM queries.

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

The paper argues that the conventional blocking-matching-clustering pipeline for dirty entity resolution creates a static sparse graph full of missing edges and noisy links, which propagates errors into final clusters. It proposes instead to treat matching and clustering as synergistic tasks that can be solved jointly by refining graph structure and propagating labels dynamically. This approach combines inexpensive signals from graph propagation with targeted expensive LLM pairwise queries, chosen via a budgeted optimization that maximizes marginal gain. A sympathetic reader would care because the method promises higher-quality entity clusters while using fewer costly LLM calls than pipelines that keep the steps separate.

Core claim

Alper integrates matching and clustering into an iterative probabilistic label propagation process over a global, evolving graph. Unlike disjoint blocking, Alper refines the graph structure and labels dynamically by adaptively integrating weak but cheap signals from graph propagation with strong but expensive LLM-based pairwise queries. The signal selection is formulated as a constrained optimization problem maximizing cumulative marginal gain under a query budget, solved via a greedy algorithm with provable theoretical guarantees.

What carries the argument

Iterative probabilistic label propagation over an evolving graph that adaptively mixes cheap propagation signals with budgeted LLM queries, solved by a greedy marginal-gain algorithm.

If this is right

  • Yields consistently higher cluster quality than state-of-the-art cascaded pipelines across eight benchmark datasets.
  • Reduces error propagation that arises from static graphs produced by decoupled blocking, matching, and clustering steps.
  • Achieves higher cost-effectiveness by selecting LLM queries that maximize cumulative marginal gain under a fixed budget.
  • Supplies theoretical guarantees that the greedy algorithm for query selection is near-optimal.
  • Produces a final entity graph whose edges and labels improve together rather than remaining fixed after early blocking decisions.

Where Pith is reading between the lines

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

  • The same adaptive combination of cheap propagation signals and selective expensive queries could be tested on incremental or streaming entity-resolution settings where the graph grows over time.
  • The framework might extend to other data-cleaning tasks such as deduplication in knowledge graphs where LLM calls are the dominant cost.
  • If the marginal-gain formulation proves robust, similar budgeted optimization could replace heuristic query selection in active-learning pipelines for graph labeling.
  • The evolving-graph view suggests that error modes introduced by early blocking decisions could be corrected later, an effect worth measuring separately from overall F1 scores.

Load-bearing premise

Matching and clustering can be jointly optimized through iterative probabilistic label propagation on an evolving graph without the integration itself introducing new error modes or instability.

What would settle it

If experiments on the eight benchmark datasets showed Alper producing clusters of equal or lower quality, or requiring equal or higher total LLM queries, than the best cascaded pipelines, the superiority claim would be falsified.

Figures

Figures reproduced from arXiv: 2605.25814 by Haoran Zheng, Hongtao Wang, Renchi Yang, Xiangyu Ke.

Figure 2
Figure 2. Figure 2: The CBMR when varying redundancy factor. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: An overview of Alper iteratively and jointly refined or optimized with complementary signals from the LLM-based matcher and graph neighborhoods. Par￾ticularly, the weighted propagation of labels via neighbors enables the use of soft transitivity to infer “easy” matches probabilistically without query cost, which not only attenuates the large errors amplified by applying rigid clustering transitivity over i… view at source ↗
Figure 4
Figure 4. Figure 4: ER performance when varying query budget [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 6
Figure 6. Figure 6: Varying 𝛼. 0.2 0.4 0.6 0.8 1 88.5 89.0 89.5 90.0 90.5 NMI (a) Cora 0.2 0.4 0.6 0.8 1 88.0 88.5 89.0 89.5 90.0 (b) Alaska 0.2 0.4 0.6 0.8 1 86.5 87.0 87.5 88.0 88.5 (c) Music 0.2 0.4 0.6 0.8 1 75.5 76.0 76.5 77.0 77.5 (d) Movies [PITH_FULL_IMAGE:figures/full_fig_p008_6.png] view at source ↗
Figure 9
Figure 9. Figure 9: An example of the prompt used for LLM verification [PITH_FULL_IMAGE:figures/full_fig_p016_9.png] view at source ↗
read the original abstract

Dirty entity resolution (ER), which identifies records referring to the same real-world entity from a single, messy dataset, is a fundamental task in data management and mining. However, the dominant blocking-matching-clustering paradigm for ER suffers from critical flaws. Its cascaded, decoupled workflow essentially produces a static, sparse graph plagued by missing edges (due to blocking failures) and noisy links (due to matching errors), causing error propagation and yielding suboptimal clusters, particularly when rigid transitivity is imposed in the clustering. We contend that matching and clustering are fundamentally synergistic, both optimizing for the construction of an ideal entity graph. Building upon this insight, we propose Alper, a unified framework that integrates these steps into an iterative probabilistic label propagation process over a global, evolving graph. Unlike disjoint blocking, Alper refines the graph structure and labels dynamically by adaptively integrating "weak but cheap" signals from graph propagation with "strong but expensive" LLM-based pairwise queries. For higher cost-effectiveness, we formulate the signal selection as a constrained optimization problem maximizing cumulative marginal gain under a query budget, solved via our greedy algorithm with provable theoretical guarantees. Our extensive experiments over eight benchmark datasets demonstrate that Alper is consistently superior to state-of-the-art cascaded pipelines.

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

3 major / 1 minor

Summary. The paper proposes Alper, a unified framework for dirty entity resolution that integrates matching and clustering via iterative probabilistic label propagation on an evolving graph. It adaptively combines cheap graph-based signals with expensive LLM pairwise queries under a query budget, formulated as a constrained optimization problem solved by a greedy algorithm claimed to have provable theoretical guarantees. The central empirical claim is that Alper is consistently superior to state-of-the-art cascaded blocking-matching-clustering pipelines across eight benchmark datasets.

Significance. If the reported superiority and the theoretical guarantees hold under rigorous evaluation, the work could meaningfully advance entity resolution by demonstrating that joint optimization of matching and clustering reduces error propagation compared to decoupled pipelines. The cost-effective use of LLMs via adaptive signal selection would be a practical contribution to hybrid graph-ML methods in data management. However, the absence of any experimental details, metrics, baselines, or derivation in the provided manuscript text prevents confirmation that the results support these implications.

major comments (3)
  1. [Abstract] Abstract: the claim of 'consistent superiority' on eight benchmarks and 'provable theoretical guarantees' for the greedy algorithm is asserted without any metrics (e.g., F1, precision-recall), baselines, statistical tests, or derivation of the optimization problem or approximation ratio; this is load-bearing for both the empirical and theoretical contributions.
  2. [unified framework description] The section on the unified framework: the premise that matching and clustering are 'fundamentally synergistic' and that iterative label propagation integrates them without introducing new error modes or instability lacks any supporting ablation (e.g., non-iterative variant) or error-propagation analysis, which directly underpins the motivation for departing from the cascaded paradigm.
  3. [signal selection optimization] The optimization formulation: no equation defining the constrained optimization, marginal gain, or query budget constraint is supplied, so it is impossible to verify whether the greedy selector's guarantees are non-vacuous or whether the 'weak but cheap' vs. 'strong but expensive' integration is correctly modeled.
minor comments (1)
  1. [Abstract] The abstract refers to 'Alper' without expanding the acronym or providing a high-level system diagram, which would aid readability.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the detailed comments. We agree that the provided manuscript text does not contain the requested equations, ablations, metrics, or derivations, and we will revise the manuscript to incorporate them.

read point-by-point responses

  1. Referee: [Abstract] Abstract: the claim of 'consistent superiority' on eight benchmarks and 'provable theoretical guarantees' for the greedy algorithm is asserted without any metrics (e.g., F1, precision-recall), baselines, statistical tests, or derivation of the optimization problem or approximation ratio; this is load-bearing for both the empirical and theoretical contributions.

    Authors: We will revise the abstract to include key quantitative results (e.g., average F1 improvements and baseline names) while preserving conciseness. Full metrics, statistical tests, and the theoretical derivation will be added to the main body and referenced from the abstract. revision: yes

  2. Referee: [unified framework description] The section on the unified framework: the premise that matching and clustering are 'fundamentally synergistic' and that iterative label propagation integrates them without introducing new error modes or instability lacks any supporting ablation (e.g., non-iterative variant) or error-propagation analysis, which directly underpins the motivation for departing from the cascaded paradigm.

    Authors: We will add an ablation comparing iterative vs. non-iterative variants and an explicit error-propagation analysis in the revised manuscript to support the unified framework motivation. revision: yes

  3. Referee: [signal selection optimization] The optimization formulation: no equation defining the constrained optimization, marginal gain, or query budget constraint is supplied, so it is impossible to verify whether the greedy selector's guarantees are non-vacuous or whether the 'weak but cheap' vs. 'strong but expensive' integration is correctly modeled.

    Authors: We will insert the formal constrained optimization equation, marginal gain definition, budget constraint, and the greedy algorithm's approximation ratio derivation in the revised Section 4. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper's derivation introduces a unified iterative label propagation framework for ER, formulates signal selection as a budgeted optimization solved by a greedy algorithm with stated theoretical guarantees, and validates via external experiments on eight benchmarks. No load-bearing step reduces by construction to fitted inputs, self-definitions, or self-citation chains; the empirical superiority claim is independent of internal parameter definitions. The framework and guarantees are presented as novel contributions without the patterns of self-definitional equivalence or renamed known results.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review; no explicit free parameters, axioms, or invented entities are stated. The framework implicitly relies on the domain assumption that entity resolution graphs admit useful probabilistic label propagation, but this is not quantified.

pith-pipeline@v0.9.1-grok · 5758 in / 1225 out tokens · 32703 ms · 2026-06-29T21:27:37.478918+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

79 extracted references · 10 canonical work pages · 3 internal anchors

  1. [1]

    Nikhil Bansal, Avrim Blum, and Shuchi Chawla. 2004. Correlation clustering. Machine learning56, 1 (2004), 89–113

    2004

  2. [2]

    Omar Benjelloun, Hector Garcia-Molina, David Menestrina, Qi Su, Steven Eui- jong Whang, and Jennifer Widom. 2009. Swoosh: a generic approach to entity resolution.The VLDB Journal18, 1 (2009), 255–276

    2009

  3. [3]

    Brenda Betancourt, Giacomo Zanella, Jeffrey W Miller, Hanna Wallach, Abbas Zaidi, and Rebecca C Steorts. 2016. Flexible models for microclustering with application to entity resolution.Advances in neural information processing systems 29 (2016)

    2016

  4. [4]

    Christoph Böhm, Gerard De Melo, Felix Naumann, and Gerhard Weikum. 2012. LINDA: distributed web-of-data-scale entity matching. InProceedings of the 21st ACM international conference on Information and knowledge management. 2104– 2108

    2012

  5. [5]

    Chengliang Chai, Guoliang Li, Jian Li, Dong Deng, and Jianhua Feng. 2016. Cost- effective crowdsourced entity resolution: A partial-order approach. InProceedings of the 2016 International Conference on Management of Data. 969–984

    2016

  6. [6]

    Deeparnab Chakrabarty, Yunhong Zhou, and Rajan Lukose. 2008. Online knap- sack problems. InWorkshop on internet and network economics (WINE). 1–9

    2008

  7. [7]

    Zhaoqi Chen, Dmitri V Kalashnikov, and Sharad Mehrotra. 2005. Exploiting relationships for object consolidation. InProceedings of the 2nd international workshop on Information quality in information systems. 47–58

    2005

  8. [8]

    Peter Christen. 2011. A survey of indexing techniques for scalable record linkage and deduplication.IEEE transactions on knowledge and data engineering24, 9 (2011), 1537–1555

    2011

  9. [9]

    Peter Christen. 2012. The data matching process. InData matching: concepts and techniques for record linkage, entity resolution, and duplicate detection. Springer, 23–35

    2012

  10. [10]

    Vassilis Christophides, Vasilis Efthymiou, Themis Palpanas, George Papadakis, and Kostas Stefanidis. 2020. An overview of end-to-end entity resolution for big data.ACM Computing Surveys (CSUR)53, 6 (2020), 1–42

    2020

  11. [11]

    Valter Crescenzi, Andrea De Angelis, Donatella Firmani, Maurizio Mazzei, Paolo Merialdo, Federico Piai, and Divesh Srivastava. 2021. Alaska: A flexible benchmark for data integration tasks.arXiv preprint arXiv:2101.11259(2021)

    work page arXiv 2021

  12. [12]

    Gianluca Demartini, Djellel Eddine Difallah, and Philippe Cudré-Mauroux. 2012. Zencrowd: leveraging probabilistic reasoning and crowdsourcing techniques for large-scale entity linking. InProceedings of the 21st international conference on World Wide Web. 469–478

    2012

  13. [13]

    Xin Dong, Evgeniy Gabrilovich, Geremy Heitz, Wilko Horn, Ni Lao, Kevin Mur- phy, Thomas Strohmann, Shaohua Sun, and Wei Zhang. 2014. Knowledge vault: A web-scale approach to probabilistic knowledge fusion. InProceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining. 601–610

    2014

  14. [14]

    Uwe Draisbach, Peter Christen, and Felix Naumann. 2019. Transforming pairwise duplicates to entity clusters for high-quality duplicate detection.Journal of Data and Information Quality (JDIQ)12, 1 (2019), 1–30

    2019

  15. [15]

    Muhammad Ebraheem, Saravanan Thirumuruganathan, Shafiq Joty, Mourad Ouzzani, and Nan Tang. 2017. DeepER–Deep Entity Resolution.arXiv preprint arXiv:1710.00597(2017)

    work page arXiv 2017

  16. [16]

    Ahmed K Elmagarmid, Panagiotis G Ipeirotis, and Vassilios S Verykios. 2006. Duplicate record detection: A survey.IEEE Transactions on knowledge and data engineering19, 1 (2006), 1–16

    2006

  17. [17]

    Meihao Fan, Xiaoyue Han, Ju Fan, Chengliang Chai, Nan Tang, Guoliang Li, and Xiaoyong Du. 2024. Cost-effective in-context learning for entity resolution: A design space exploration. In2024 IEEE 40th International Conference on Data Engineering (ICDE). IEEE, 3696–3709

    2024

  18. [18]

    Ivan P Fellegi and Alan B Sunter. 1969. A theory for record linkage.Journal of the American statistical association64, 328 (1969), 1183–1210

    1969

  19. [19]

    Jiajie Fu, Haitong Tang, Arijit Khan, Sharad Mehrotra, Xiangyu Ke, and Yunjun Gao. 2025. In-context clustering-based entity resolution with large language models: A design space exploration.Proceedings of the ACM on Management of Data3, 4 (2025), 1–28

    2025

  20. [20]

    Sainyam Galhotra, Donatella Firmani, Barna Saha, and Divesh Srivastava. 2018. Robust entity resolution using random graphs. InProceedings of the 2018 Interna- tional Conference on Management of Data. 3–18

    2018

  21. [21]

    Congcong Ge, Pengfei Wang, Lu Chen, Xiaoze Liu, Baihua Zheng, and Yunjun Gao. 2021. CollaborEM: A self-supervised entity matching framework using multi- features collaboration.IEEE Transactions on Knowledge and Data Engineering35, 12 (2021), 12139–12152

    2021

  22. [22]

    Oktie Hassanzadeh, Fei Chiang, Hyun Chul Lee, and Renée J Miller. 2009. Frame- work for evaluating clustering algorithms in duplicate detection.Proceedings of the VLDB Endowment2, 1 (2009), 1282–1293

    2009

  23. [23]

    Mauricio A Hernández and Salvatore J Stolfo. 1995. The merge/purge problem for large databases.ACM Sigmod Record24, 2 (1995), 127–138

    1995

  24. [24]

    Jeff Howe. 2006. The rise of crowdsourcing, Wired.http://www. wired. com/wired/archive/14.06/crowds. html(2006)

    2006

  25. [25]

    Keke Huang, Yimin Shi, Dujian Ding, Yifei Li, Yang Fei, Laks Lakshmanan, and Xiaokui Xiao. 2025. Thriftllm: On cost-effective selection of large language models for classification queries.arXiv preprint arXiv:2501.04901(2025)

    work page arXiv 2025

  26. [26]

    Anitha Kannan, Inmar E Givoni, Rakesh Agrawal, and Ariel Fuxman. 2011. Match- ing unstructured product offers to structured product specifications. InProceed- ings of the 17th ACM SIGKDD international conference on Knowledge discovery and data mining. 404–412

    2011

  27. [27]

    Pradap Konda, Sanjib Das, Paul Suganthan G C, AnHai Doan, Adel Ardalan, Jeffrey R Ballard, Han Li, Fatemah Panahi, Haojun Zhang, Jeff Naughton, et al

  28. [28]

    Magellan: toward building entity matching management systems over data science stacks.Proceedings of the VLDB Endowment9, 13 (2016), 1581–1584

    2016

  29. [29]

    Simon Lacoste-Julien, Konstantina Palla, Alex Davies, Gjergji Kasneci, Thore Graepel, and Zoubin Ghahramani. 2013. Sigma: Simple greedy matching for align- ing large knowledge bases. InProceedings of the 19th ACM SIGKDD international conference on Knowledge discovery and data mining. 572–580

    2013

  30. [30]

    Bing Li, Wei Wang, Yifang Sun, Linhan Zhang, Muhammad Asif Ali, and Yi Wang

  31. [31]

    InProceedings of the AAAI Conference on Artificial Intelligence, Vol

    Grapher: Token-centric entity resolution with graph convolutional neural networks. InProceedings of the AAAI Conference on Artificial Intelligence, Vol. 34. 8172–8179

  32. [32]

    Huahang Li, Longyu Feng, Shuangyin Li, Fei Hao, Chen Jason Zhang, and Yuan- feng Song. 2024. On leveraging large language models for enhancing entity resolution: a cost-efficient approach.arXiv preprint arXiv:2401.03426(2024)

    work page arXiv 2024

  33. [33]

    Yuliang Li, Jinfeng Li, Yoshihiko Suhara, AnHai Doan, and Wang-Chiew Tan

  34. [34]

    Deep entity matching with pre-trained language models.arXiv preprint arXiv:2004.00584(2020)

    work page arXiv 2004

  35. [35]

    Andrew McCallum, Kamal Nigam, and Lyle H Ungar. 2000. Efficient clustering of high-dimensional data sets with application to reference matching. InProceedings of the sixth ACM SIGKDD international conference on Knowledge discovery and data mining. 169–178

    2000

  36. [36]

    Venkata Vamsikrishna Meduri, Lucian Popa, Prithviraj Sen, and Mohamed Sarwat

  37. [37]

    InProceedings of the 2020 ACM SIGMOD international conference on management of data

    A comprehensive benchmark framework for active learning methods in entity matching. InProceedings of the 2020 ACM SIGMOD international conference on management of data. 1133–1147

    2020

  38. [38]

    Barzan Mozafari, Purna Sarkar, Michael Franklin, Michael Jordan, and Samuel Madden. 2014. Scaling up crowd-sourcing to very large datasets: a case for active learning.Proceedings of the VLDB Endowment8, 2 (2014), 125–136

    2014

  39. [39]

    Avanika Narayan, Ines Chami, Laurel Orr, Simran Arora, and Christopher Ré

  40. [40]

    Can foundation models wrangle your data?arXiv preprint arXiv:2205.09911 (2022)

    work page arXiv 2022

  41. [41]

    Mahdi Niknam, Behrouz Minaei-Bidgoli, and Rouhollah Dianat. 2022. The role of transitive closure in evaluating blocking methods for dirty entity resolution. Journal of Intelligent Information Systems58, 3 (2022), 561–590

    2022

  42. [42]

    Konstantinos Nikoletos, George Papadakis, and Manolis Koubarakis. 2022. py- JedAI: a Lightsaber for Link Discovery.. InISWC (Posters/Demos/Industry)

    2022

  43. [43]

    George Papadakis, Dimitrios Skoutas, Emmanouil Thanos, and Themis Palpanas

  44. [44]

    9 Conference’17, July 2017, Washington, DC, USA Wang et al

    Blocking and filtering techniques for entity resolution: A survey.ACM Computing Surveys (CSUR)53, 2 (2020), 1–42. 9 Conference’17, July 2017, Washington, DC, USA Wang et al

    2020

  45. [45]

    George Papadakis, Leonidas Tsekouras, Emmanouil Thanos, George Gian- nakopoulos, Themis Palpanas, and Manolis Koubarakis. 2018. The return of jedai: End-to-end entity resolution for structured and semi-structured data.Pro- ceedings of the VLDB Endowment, Vol. 11, No. 1211, 12 (2018), 1950–1953

    2018

  46. [46]

    Derek Paulsen, Yash Govind, and AnHai Doan. 2023. Sparkly: A simple yet surprisingly strong TF/IDF blocker for entity matching.Proceedings of the VLDB Endowment16, 6 (2023), 1507–1519

    2023

  47. [47]

    Ralph Peeters and Christian Bizer. 2021. Dual-objective fine-tuning of BERT for entity matching.Proceedings of the VLDB Endowment14 (2021), 1913–1921

    2021

  48. [48]

    Clifton Phua, Vincent Lee, Kate Smith, and Ross Gayler. 2010. A compre- hensive survey of data mining-based fraud detection research.arXiv preprint arXiv:1009.6119(2010)

    work page internal anchor Pith review Pith/arXiv arXiv 2010

  49. [49]

    Usha Nandini Raghavan, Réka Albert, and Soundar Kumara. 2007. Near linear time algorithm to detect community structures in large-scale networks.Physical Review E—Statistical, Nonlinear, and Soft Matter Physics76, 3 (2007), 036106

    2007

  50. [50]

    Alieh Saeedi, Markus Nentwig, Eric Peukert, and Erhard Rahm. 2018. Scalable matching and clustering of entities with FAMER.Complex Systems Informatics and Modeling Quarterly16 (2018), 61–83

    2018

  51. [51]

    Alieh Saeedi, Eric Peukert, and Erhard Rahm. 2017. Comparative evaluation of distributed clustering schemes for multi-source entity resolution. InEuropean Conference on Advances in Databases and Information Systems. Springer, 278–293

    2017

  52. [52]

    Claude E Shannon. 1948. A mathematical theory of communication.The Bell system technical journal27, 3 (1948), 379–423

    1948

  53. [53]

    Fabian M Suchanek, Serge Abiteboul, and Pierre Senellart. 2011. Paris: Probabilis- tic alignment of relations, instances, and schema.arXiv preprint arXiv:1111.7164 (2011)

    work page internal anchor Pith review Pith/arXiv arXiv 2011

  54. [54]

    Saravanan Thirumuruganathan, Han Li, Nan Tang, Mourad Ouzzani, Yash Govind, Derek Paulsen, Glenn Fung, and AnHai Doan. 2021. Deep learning for blocking in entity matching: a design space exploration.Proceedings of the VLDB Endowment 14, 11 (2021), 2459–2472

    2021

  55. [55]

    Samhita Vadrevu, Rakesh Nagi, JinJun Xiong, and Wen-mei Hwu. 2021. xER: An explainable model for entity resolution using an efficient solution for the clique partitioning problem. InProceedings of the First Workshop on Trustworthy Natural Language Processing. 34–44

    2021

  56. [56]

    Vasilis Verroios and Hector Garcia-Molina. 2015. Entity resolution with crowd errors. In2015 IEEE 31st international conference on data engineering. IEEE, 219– 230

    2015

  57. [57]

    Vasilis Verroios, Hector Garcia-Molina, and Yannis Papakonstantinou. 2017. Waldo: An adaptive human interface for crowd entity resolution. InProceedings of the 2017 ACM International Conference on Management of Data. 1133–1148

    2017

  58. [58]

    Norases Vesdapunt, Kedar Bellare, and Nilesh Dalvi. 2014. Crowdsourcing algo- rithms for entity resolution.Proceedings of the VLDB Endowment7, 12 (2014), 1071–1082

    2014

  59. [59]

    Jiannan Wang, Tim Kraska, Michael J Franklin, and Jianhua Feng. 2012. Crowder: Crowdsourcing entity resolution.arXiv preprint arXiv:1208.1927(2012)

    work page internal anchor Pith review Pith/arXiv arXiv 2012

  60. [60]

    Jiannan Wang, Guoliang Li, Tim Kraska, Michael J Franklin, and Jianhua Feng

  61. [61]

    InProceedings of the 2013 ACM SIGMOD International Conference on Management of Data

    Leveraging transitive relations for crowdsourced joins. InProceedings of the 2013 ACM SIGMOD International Conference on Management of Data. 229–240

    2013

  62. [62]

    Sibo Wang, Xiaokui Xiao, and Chun-Hee Lee. 2015. Crowd-based deduplication: An adaptive approach. InProceedings of the 2015 ACM SIGMOD international conference on Management of Data. 1263–1277

    2015

  63. [63]

    Tianshu Wang, Xiaoyang Chen, Hongyu Lin, Xuanang Chen, Xianpei Han, Le Sun, Hao Wang, and Zhenyu Zeng. 2025. Match, compare, or select? an investigation of large language models for entity matching. InProceedings of the 31st International Conference on Computational Linguistics. 96–109

    2025

  64. [64]

    Yaoshu Wang, Jianbin Qin, and Wei Wang. 2017. Efficient approximate entity matching using jaro-winkler distance. InInternational conference on web infor- mation systems engineering. Springer, 231–239

    2017

  65. [65]

    Steven Euijong Whang, Peter Lofgren, and Hector Garcia-Molina. 2013. Question selection for crowd entity resolution. (2013)

    2013

  66. [66]

    William E Winkler et al. 2006. Overview of record linkage and current research directions.Bureau of the Census25, 4 (2006), 603–623

    2006

  67. [67]

    Renzhi Wu, Alexander Bendeck, Xu Chu, and Yeye He. 2023. Ground truth inference for weakly supervised entity matching.Proceedings of the ACM on Management of Data1, 1 (2023), 1–28

    2023

  68. [68]

    Renzhi Wu, Sanya Chaba, Saurabh Sawlani, Xu Chu, and Saravanan Thirumu- ruganathan. 2020. Zeroer: Entity resolution using zero labeled examples. In Proceedings of the 2020 ACM SIGMOD International Conference on Management of Data. 1149–1164

    2020

  69. [69]

    Vijaya Krishna Yalavarthi, Xiangyu Ke, and Arijit Khan. 2017. Select your ques- tions wisely: For entity resolution with crowd errors. InProceedings of the 2017 ACM on Conference on Information and Knowledge Management. 317–326

    2017

  70. [70]

    Dezhong Yao, Yuhong Gu, Gao Cong, Hai Jin, and Xinqiao Lv. 2022. Entity resolution with hierarchical graph attention networks. InProceedings of the 2022 International Conference on Management of Data. 429–442

    2022

  71. [71]

    Li Yujian and Liu Bo. 2007. A normalized Levenshtein distance metric.IEEE transactions on pattern analysis and machine intelligence29, 6 (2007), 1091–1095

    2007

  72. [72]

    control queries

    Haochen Zhang, Yuyang Dong, Chuan Xiao, and Masafumi Oyamada. 2023. Large language models as data preprocessors.arXiv preprint arXiv:2308.16361(2023). 10 Adaptive Graph Refinement and Label Propagation with LLMs for Cost-Effective Entity Resolution Conference’17, July 2017, Washington, DC, USA A Additional Related Work In this section, we provide suppleme...

    work page arXiv 2023

  73. [73]

    Please respond with ONLY the number (1, 2, ...) of the matching candidate, or "NONE" if no match is found

  74. [74]

    Input Prompt Query Entity: Intermezzo in B minor, Op

    Your response should be a single number or "NONE", nothing else. Input Prompt Query Entity: Intermezzo in B minor, Op. 119 No. 1: Adago - Brahms: Complete Works Candidate Entities:

  75. [75]

    12 in B minor (HWV 330) - I

    Opus 6 No. 12 in B minor (HWV 330) - I. Largo - Concerti Grossi op. 6

  76. [76]

    003-Symphony 1 in C minor, op. 11: III. Menuetto & Trio, Allegro di Molto

  77. [77]

    Intermezzo in B minor, Op. 119 No. 1: Adagio

  78. [78]

    017-Intermezzo in B minor, Op. 119 No. 1: Adagio

  79. [79]

    Johannes Brahms - Concerto for Violin and Orchestra in D major, Op. 77: II. Adagio Response 3 Figure 9: An example of the prompt used for LLM verification onMusic. C.13 Analysis of Error Correction A key advantage of Alper is its ability to mitigate early-stage LLM errors. “Cheap” LLMs risk generating false positives, but our frame- work absorbs these thr...