Pith. sign in

REVIEW 5 major objections 6 minor 39 references

Leveraging Large Language Models for Generating Labeled Mineral Site Record Linkage Data

T0 review · 5 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read LLM-generated labels can substitute for expert-curated ground truth in mineral site record linkage, yielding a 45 percent F1 gain over a ground-truth-trained model.

desk verdict The headline result is an artifact of a degenerate baseline; the paper is a useful distillation study whose evaluation needs a non-degenerate GT baseline and larger test sets before its central claim is credible. read the letter →

arxiv 2412.03575 v1 pith:BVBSAZLJ submitted 2024-11-17 cs.IR cs.AIcs.CL

classification cs.IRcs.AIcs.CL
keywords recordlinkageentityresolutionspatiallargelanguagemodelspretrainedLLM-generatedtrainingdatamineralsiteimbalance
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

Where human-curated labels for linking mineral site records are scarce, the paper argues that labels generated by a large language model can fill the gap. It proposes a two-stage pipeline: LLaMA3-8b labels random record pairs as matches or non-matches, and those labels fine-tune RoBERTa, a small pretrained encoder, for binary classification. On tungsten and nickel test sets, the resulting model raises macro-averaged F1 by more than 45 percent relative to a RoBERTa model trained on the available ground truth, while cutting inference time by a factor of about 18 compared with using the LLM itself. The importance is that an automated, human-free pipeline can make record linkage practical for large mineral databases.

What carries the argument

The machinery is a two-stage transfer of matching semantics. In the first stage, record pairs are serialized into the text template "Entity A is ... Entity B is ..." and LLaMA3-8b is forced to answer "Yes" or "No" about whether they describe the same mine, converting its answers into binary labels. In the second stage, the same pairs are re-serialized with special tokens ([COL] for attribute names, [VAL] for values, [CLS]/[SEP] for sequence structure) and used to fine-tune RoBERTa as a match/non-match classifier. The choice that carries the argument is to spend the LLM's expensive inference only once, at data-generation time, and deploy the fast PLM for every future prediction.

What would settle it

Compute the set intersection between the 80 percent of LLaMA-labeled training pairs and the 10 percent tungsten ground-truth test pairs; any shared record pair would make the F1 gain possibly a memorization effect. Then rerun the tungsten evaluation with a larger test split containing dozens of match pairs instead of two or three, and retrain the ground-truth baseline on a class-balanced sample of the available matches; if the baseline's F1 closes the gap, the reported 45 percent improvement depends on the baseline's collapse to the majority class.

Watch

Extended reading notes

Core claim

The paper's central claim is that LLM-generated training labels can substitute for expert-curated ground truth in mineral site record linkage, and that a small classifier trained on those labels inherits most of the LLM's matching ability. Using a yes/no prompting protocol, LLaMA3-8b labels serialized record pairs; those labels fine-tune RoBERTa. On the tungsten dataset the fine-tuned model matches the zero-shot LLM's macro-averaged F1 of 73.05 percent, and on the nickel dataset it reaches 77.43 percent versus 88.11 percent for the LLM, while the ground-truth-trained PLM manages only 48.31 percent. The authors interpret the result as evidence that the LLM's semantic understanding of attributes and location transfers to the efficient PLM through the synthetic labels, making the approach a viable alternative to costly manual labeling.

Load-bearing premise

The entire comparison rests on the assumption that the LLaMA-labeled training pairs and the ground-truth test pairs are disjoint, and that the test sets hold enough match pairs for the reported F1 numbers to be statistically meaningful.

Editorial extensions

If this is right

  • Mineral site record linkage can be run without expert-curated training labels, replacing them with automatically generated LLM labels.
  • The fine-tuned PLM delivers the LLM's match quality at roughly one-eighteenth the inference time, making linkages of databases with hundreds of thousands of records plausible.
  • The method transfers across commodities and regions: a model trained on data spanning diverse minerals achieves competitive F1 on the separate tungsten and nickel test sets.
  • The data-size experiments suggest a practical rule: around 250 match samples in the LLM-labeled training set are enough for the match F1 to stabilize.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The same generate-and-fine-tune recipe could extend to other domains where labeled tabular data is scarce, such as linking scientific catalogs or historical records, since the pipeline makes no mineral-specific assumptions beyond the prompt wording.
  • Because the LLM's own tungsten match F1 is only about 39 percent, the PLM's success suggests that abundant but noisy synthetic labels can outperform a few clean but extremely imbalanced ground-truth labels; balancing the ground-truth baseline would clarify how much of the gain is simply correcting class imbalance.
  • A natural extension would be active or iterative labeling: have the LLM label only pairs the PLM is uncertain about, which would cut labeling cost while adding match examples to the training set.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 6 minor

Summary. The paper proposes a two-stage method for mineral site record linkage: use LLaMA3-8b to generate binary match/non-match labels for record pairs from the MRDS and USMIN databases, then fine-tune a RoBERTa PLM on those labels. The authors report that the resulting model improves macro-averaged F1 by over 45% over a RoBERTa model trained on expert ground truth (GT-Trained) and that inference is nearly 18 times faster than using LLaMA alone. Evaluation uses two USGS curated benchmarks, the Tungsten and Nickel datasets, plus manually curated, GeoER, and LLaMA-only baselines. The paper also presents experiments on training-data size and class imbalance.

Significance. If substantiated, the central idea has clear value: LLM-generated labels could replace expensive expert-curated training data for domain-specific record linkage, and the PLM student would provide a large inference-time advantage over LLM-only approaches. The application domain (mineral site data) is concrete and under-served, and the paper uses two real-world ground-truth datasets. The data-size and imbalance analyses (Figures 8-10) are a useful practical contribution, and the authors include a thoughtful ethical-consideration discussion of hallucination risks and energy costs. However, the quantitative headline claims are not currently supported: the main baseline comparison is against a degenerate classifier that predicts all non-matches, the positive test pool is extremely small, and a record-level train/test overlap is not ruled out.

major comments (5)
  1. [§4.1.5 and Table 4] The GT-Trained baseline is degenerate: the 80% training split of the Tungsten ground truth contains only about 18 of the 23 positive pairs, and the manuscript itself states that this model predicts 'No' for all pairs, giving match F1=0.00 and macro F1≈49.99 on Tungsten and 48.31 on Nickel. The abstract's 'over 45% improvement' is computed against this collapsed baseline, so it does not demonstrate that LLM labels substitute for expert labels. The authors must add a functioning ground-truth baseline, e.g., training with all 23 positives via cross-validation or with class weighting/oversampling, before the central claim can be credited.
  2. [§4.1.4 and Table 4] The Tungsten test set is 10% of the ground truth, which is only 2-3 positive pairs (out of 23). The reported match F1 values (46.15 for both LLaMA3-8b and the proposed method) are therefore based on essentially no positive evidence, and differences between models are not statistically meaningful. The paper should report the number of true positives in each test split, provide error bars or confidence intervals (e.g., via repeated holdout), or use a larger evaluation setup such as k-fold cross-validation over all ground-truth pairs.
  3. [§4.1.1 and §4.1.4] The LLaMA-labeled training data is generated by randomly selecting records from the same MRDS and USMIN databases that constitute the Tungsten and Nickel ground-truth test data, and the paper does not state that records in the test pairs were excluded from training. Since §4.1.4 splits only the LLaMA-labeled pairs and not the ground-truth records, record-level leakage is possible: if a record appearing in a test pair also appears in a training pair, the model could memorize it, inflating the reported F1 scores. The authors must clarify whether test records were excluded from training-data generation or quantify the overlap.
  4. [Abstract, §4.2 and Table 4] The headline claim of 'over 45% improvement in F1 score compared to traditional PLM-based methods' is only supported against the degenerate GT-Trained baseline. On the Nickel dataset the proposed method is worse than the LLaMA3-8b baseline (match F1 57.14 vs 77.78; macro F1 77.43 vs 88.11), and the only other PLM-based baseline, GeoER, is not comparable on Tungsten because it did not finish. The abstract and conclusions should be revised to describe the actual comparison set and to acknowledge the quality tradeoff on Nickel, or provide a combined analysis that justifies a general claim.
  5. [Appendix C, §4.2.2] The 'nearly 18 times' inference-time reduction relies on extrapolated quadratic fits with coefficients 0.073 and 0.004, but the paper does not report the raw measured times, the number of measurement points, or any goodness-of-fit statistic. The extrapolation to 300,000 records is therefore not validated. The authors should include the measured times (or a table/figure with the fitted curves and residuals) and describe the measurement conditions, such as batch size and sequence length, and whether the LLaMA time includes input serialization.
minor comments (6)
  1. [§4.1.4] The validation set for the proposed approach is not specified; the text says the best epoch is chosen 'as determined by the validation data,' but it is unclear whether this validation split comes from the LLaMA-labeled data or from the ground-truth data. Please clarify.
  2. [§4.1.1 and Table 2] The sentence 'We ensure that the volume of data–387 records and 74,691 record pairs–is identical to that of the Tungsten assessment data' is confusing because the number of pairs for 387 records is determined by the combinatorial count; if the records are randomly selected, the resulting number of records and pairs should be reported as a consequence, not a target.
  3. [§4.2.3 and Figures 8-10] The evaluation dataset for the data-size and imbalance experiments is not precisely stated: it says the model is evaluated on the 'Tungsten ground truth dataset,' but it is unclear whether this is the full ground truth or a held-out split and whether the reported values are averages over multiple runs. Please specify.
  4. [§4.2.2 and Figure 7] The y-axis of Figure 7 is stated as inference time in minutes in the text, but the figure caption omits units; please add clear axis labels and units to the figure.
  5. [Appendix B vs §4.1.4] The weight decay is reported as 0.015 in Section 4.1.4 and 0.018 in Appendix B for the model comparison; these should be reconciled.
  6. [References] References [28] and [29] are duplicates (both are Reimers and Gurevich 2019), and the author list for reference [38] appears incomplete and incorrectly ordered ('Liu Zhuang' should be 'Yinhan Liu, Myle Ott, et al.' for RoBERTa); please correct the bibliography.

Circularity Check

0 steps flagged · score 0.0 of 10

No circular derivation: the PLM is trained on LLaMA-generated labels and tested on independently curated USGS ground truth; the inflated >45% headline reflects a degenerate GT-Trained baseline, an evaluation-validity issue rather than a circularity.

full rationale

The paper's central chain is: LLaMA3-8b labels candidate record pairs; those labels fine-tune RoBERTa; RoBERTa is evaluated against expert-curated Tungsten and Nickel ground truth (Goldman et al.; USGS expert Zientek). The test labels therefore do not come from the model or from the LLM that produced the training labels, so the F1 numbers are not defined in terms of the method's own outputs. The training and test pairs are not explicitly shown to be disjoint, and the GT-Trained baseline's match F1 of 0.00 arises from a split that leaves roughly 18 positive training pairs, below the 25-example threshold the paper itself cites from Tänzer et al. These are experimental-validity weaknesses, not circular reductions: no reported equation is equivalent to its input by construction, and no load-bearing claim rests on the authors' own prior results. The only self-citation (SpaBERT, [19]) appears in a future-work suggestion for spatial embeddings and does not support any reported result. The inference-time extrapolations in Appendix C are explicitly fitted to measured runtimes and presented as approximations, not as independent predictions. Accordingly, no circularity step can be exhibited.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

The central claim rests on the accuracy of LLM-generated labels and the sufficiency of the tiny test sets. No new entities are postulated. The only fitted values are the runtime extrapolation coefficients, which support the speed claim but not the accuracy claim.

free parameters (2)
  • LLaMA inference time coefficient = 0.073
    Fitted to measured inference times in Appendix C to extrapolate runtime at 300,000 records.
  • Proposed method inference time coefficient = 0.004
    Fitted to measured inference times in Appendix C; used to claim an 18x speedup.
assumptions (4)
  • domain assumption The Tungsten and Nickel ground truth labels from USGS experts are correct.
    Used as test labels in Table 4; any ground truth error directly changes the reported F1.
  • domain assumption LLaMA3-8b zero-shot labels are a reliable enough proxy for true matches to train a downstream classifier.
    This is the core premise of the method; Section 4.1.3 validates LLaMA's accuracy but does not audit the errors that the PLM will inherit.
  • domain assumption Randomly sampling records from MRDS and USMIN produces training pairs representative of the Tungsten and Nickel test distributions.
    The authors sample nationwide but test on two regional subsets; no distributional analysis is provided.
  • domain assumption RoBERTa and LLaMA3-8b pretrained weights are used as-is without modification.
    The method relies on pretrained models retaining their capabilities on the serialized mineral data.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Leveraging Large Language Models for Generating Labeled Mineral Site Record Linkage Data." pith.science (2026). https://pith.science/paper/BVBSAZLJ

@misc{pith2026241203575,
  author       = {Pith},
  title        = {Pith review of: Leveraging Large Language Models for Generating Labeled Mineral Site Record Linkage Data},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/BVBSAZLJ}},
  note         = {Machine review of arXiv:2412.03575}
}
read the original abstract

Record linkage integrates diverse data sources by identifying records that refer to the same entity. In the context of mineral site records, accurate record linkage is crucial for identifying and mapping mineral deposits. Properly linking records that refer to the same mineral deposit helps define the spatial coverage of mineral areas, benefiting resource identification and site data archiving. Mineral site record linkage falls under the spatial record linkage category since the records contain information about the physical locations and non-spatial attributes in a tabular format. The task is particularly challenging due to the heterogeneity and vast scale of the data. While prior research employs pre-trained discriminative language models (PLMs) on spatial entity linkage, they often require substantial amounts of curated ground-truth data for fine-tuning. Gathering and creating ground truth data is both time-consuming and costly. Therefore, such approaches are not always feasible in real-world scenarios where gold-standard data are unavailable. Although large generative language models (LLMs) have shown promising results in various natural language processing tasks, including record linkage, their high inference time and resource demand present challenges. We propose a method that leverages an LLM to generate training data and fine-tune a PLM to address the training data gap while preserving the efficiency of PLMs. Our approach achieves over 45\% improvement in F1 score for record linkage compared to traditional PLM-based methods using ground truth data while reducing the inference time by nearly 18 times compared to relying on LLMs. Additionally, we offer an automated pipeline that eliminates the need for human intervention, highlighting this approach's potential to overcome record linkage challenges.

Figures

Figures reproduced from arXiv: 2412.03575 by the authors.

Figure 1
Figure 1. Illustration of the mineral site record linkage process. The pipeline must accurately link records despite variations in [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Image of Eagle Mine, where each color represents a mineral site record from different databases [ [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Image of General Washington Placer and Henderson Mine, with mineral records in different colors to represent [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Record-to-record distance distribution of match data from OSM-FSQ/OSM-Yelp compared to mineral sites data. While [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Sample of MRDS Data displaying the heterogeneity of the data. Some of the attributes are left blank, and some [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: Image illustrating true positive match records identified by our proposed approach. Each color represents a ground [PITH_FULL_IMAGE:figures/full_fig_p009_6.png]
Figure 7
Figure 7. Figure 7: Inference time (in minutes) for our approach com [PITH_FULL_IMAGE:figures/full_fig_p009_7.png]
Figure 9
Figure 9. Figure 9: Match, non-match, and macro-averaged F1 scores of [PITH_FULL_IMAGE:figures/full_fig_p010_9.png]
Figure 10
Figure 10. Figure 10: Match, non-match, and macro-averaged F1 scores [PITH_FULL_IMAGE:figures/full_fig_p010_10.png]
Figure 11
Figure 11. Figure 11: Map of the Duluth Complex. Image sourced from Bergstrom and Clarke-Sather [3]. [PITH_FULL_IMAGE:figures/full_fig_p013_11.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 15 canonical work pages

  1. [1]

    Pasquale Balsebre, Dezhong Yao, Gao Cong, and Zhen Hai. 2022. Geospatial Entity Resolution. In Proceedings of the ACM Web Conference 2022 (Virtual Event, Lyon, France) (WWW ’22). Association for Computing Machinery, New York, NY, USA, 3061–3070. https://doi.org/10.1145/3485447.3512026

  2. [2]

    Nelly Barret, Fabien Duchateau, Franck Favetta, and Ludovic Moncla. 2019. Spa- tial Entity Matching with GeoAlign (demo paper). InProceedings of the 27th ACM SIGSPATIAL International Conference on Advances in Geographic Information Sys- tems (Chicago, IL, USA) (SIGSPATIAL ’19). Association for Computing Machinery, New York, NY, USA, 580–583. https://doi.o...

  3. [3]

    Bergstrom and Afton Clarke-Sather

    Ryan D. Bergstrom and Afton Clarke-Sather. 2020. Balancing Socio-Ecological Risks, Politics, and Identity: Sustainability in Minnesota’s Copper-Nickel- Precious Metal Mining Debate. Sustainability 12, 24 (2020). https://doi.org/10. 3390/su122410286

  4. [4]

    Vadim Borisov, Kathrin Sessler, Tobias Leemann, Martin Pawelczyk, and Gjergji Kasneci. 2023. Language Models are Realistic Tabular Data Generators. In The Eleventh International Conference on Learning Representations . https:// openreview.net/forum?id=cEygmQNOeI

  5. [5]

    Tom B. Brown, Benjamin Mann, Nick Ryder, Melanie Subbiah, Jared Kaplan, Prafulla Dhariwal, Arvind Neelakantan, Pranav Shyam, Girish Sastry, Amanda Askell, Sandhini Agarwal, Ariel Herbert-Voss, Gretchen Krueger, Tom Henighan, Rewon Child, Aditya Ramesh, Daniel M. Ziegler, Jeffrey Wu, Clemens Winter, Jiyoon Pyo and Yao-Yi Chiang Christopher Hesse, Mark Chen...

  6. [6]

    Curtis Clarke, Brian Thomas, James McDonald, Stephan Blaho, Ewald Pengel, Jason Obermeyer, Devin Castendyk, and Ibrahim Karajeh. 2022. NI 43-101 Tech- nical Report on the Eagle Mine, Michigan, USA . Technical Report. WSP Canada Inc. 306 pages

  7. [7]

    Jacob Devlin, Ming-Wei Chang, Kenton Lee, and Kristina Toutanova. 2019. BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding. In Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, Volume 1 (Long and Short Papers), Jill Burstein, Christy...

  8. [8]

    Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ah- mad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, An- gela Fan, Anirudh Goyal, Anthony Hartshorn, Aobo Yang, Archi Mitra, Archie Sravankumar, Artem Korenev, Arthur Hinsvark, Arun Rao, ..., and Zhiwei Zhao. 2024. The Llama 3 Herd of Models. arXiv:2407.21783 [cs.AI] http...

Show all 39 references
  1. [9]

    Muhammad Ebraheem, Saravanan Thirumuruganathan, Shafiq Joty, Mourad Ouzzani, and Nan Tang. 2018. Distributed representations of tuples for entity resolution. Proc. VLDB Endow. 11, 11 (jul 2018), 1454–1467. https://doi.org/10. 14778/3236187.3236198

  2. [10]

    Margaret A Goldman, Connie Dicken, Brown II, Philip J, Allen K Andersen, Mitchell M Bennett, and Heather L Parks. 2022. Spatial data associated with tungsten skarn resource assessment of the Northern Rocky Mountains, Montana and Idaho. https://doi.org/10.5066/P9094RVV

  3. [11]

    Pengcheng He, Jianfeng Gao, and Weizhu Chen. 2023. DeBERTaV3: Improv- ing DeBERTa using ELECTRA-Style Pre-Training with Gradient-Disentangled Embedding Sharing. In The Eleventh International Conference on Learning Repre- sentations. https://openreview.net/forum?id=sE7-XhLxHA

  4. [12]

    Sepp Hochreiter and Jürgen Schmidhuber. 1997. Long Short-Term Memory. Neural Comput. 9, 8 (nov 1997), 1735–1780. https://doi.org/10.1162/neco.1997.9. 8.1735

  5. [13]

    Suela Isaj, Esteban Zimányi, and Torben Bach Pedersen. 2019. Multi-Source Spatial Entity Linkage. In Proceedings of the 16th International Symposium on Spatial and Temporal Databases (Vienna, Austria) (SSTD ’19). Association for Computing Machinery, New York, NY, USA, 1–10. ht...

  6. [14]

    Delaram Javdani, Hossein Rahmani, Milad Allahgholi, and Fatemeh Karimkhani

  7. [15]

    Nick A Karl, Thomas R Carroll, Meredith H Burger, Liam D Knudsen, Keith R Long, Tyler A Reyes, and German Schmeda. 2020. Tungsten Deposits in the United States (ver. 2.0, August 2020). https://doi.org/10.5066/P97NJLI4

  8. [16]

    C., AnHai Doan, Adel Ardalan, Jeffrey R

    Pradap Konda, Sanjib Das, Paul Suganthan G. C., AnHai Doan, Adel Ardalan, Jeffrey R. Ballard, Han Li, Fatemah Panahi, Haojun Zhang, Jeff Naughton, Shishir Prasad, Ganesh Krishnan, Rohit Deep, and Vijay Raghavendra. 2016. Magellan: toward building entity matching management sys...

  9. [17]

    Lawley, Anne E

    Christopher J.M. Lawley, Anne E. McCafferty, Garth E. Graham, David L. Huston, Karen D. Kelley, Karol Czarnota, Suzanne Paradis, Jan M. Peter, Nathan Hayward, Mike Barlow, Poul Emsbo, Joshua Coyan, Carma A. San Juan, and Michael G. Gadd. 2022. Data–driven prospectivity modelli...

  10. [18]

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

  11. [19]

    Zekun Li, Jina Kim, Yao-Yi Chiang, and Muhao Chen. 2022. SpaBERT: A Pretrained Language Model from Geographic Data for Geo-Entity Represen- tation. In Findings of the Association for Computational Linguistics: EMNLP 2022, Yoav Goldberg, Zornitsa Kozareva, and Yue Zhang (Eds.)....

  12. [20]

    Mason and R.E

    G.T. Mason and R.E. Arndt. 1996. Mineral Resources Data System (MRDS) . https: //doi.org/10.3133/ds20

  13. [21]

    Miller and E.M

    J.D. Miller and E.M. Ripley. 1996. Layered Intrusions of the Duluth Complex, Minnesota, USA. In Layered Intrusions, Richard Grant Cawthorn (Ed.). Develop- ments in Petrology, Vol. 15. Elsevier, 257–301. https://doi.org/10.1016/S0167- 2894(96)80010-8

  14. [22]

    Mudd and Simon M

    Gavin M. Mudd and Simon M. Jowitt. 2022. The New Century for Nickel Re- sources, Reserves, and Mining: Reassessing the Sustainability of the Devil’s Metal. Economic Geology 117, 8 (12 2022), 1961–1983. https://doi.org/10.5382/econgeo. 4950 arXiv:https://pubs.geoscienceworld.or...

  15. [23]

    Avanika Narayan, Ines Chami, Laurel Orr, and Christopher Ré. 2022. Can Foun- dation Models Wrangle Your Data? Proc. VLDB Endow. 16, 4 (dec 2022), 738–746. https://doi.org/10.14778/3574245.3574258

  16. [24]

    OpenStreetMap contributors. 2017. Planet dump retrieved from https://planet.osm.org . https://www.openstreetmap.org

  17. [25]

    Ralph Peeters and Christian Bizer. 2023. Using ChatGPT for Entity Matching. In New Trends in Database and Information Systems , Alberto Abelló, Panos Vas- siliadis, Oscar Romero, Robert Wrembel, Francesca Bugiotti, Johann Gamper, Genoveva Vargas Solar, and Ester Zumpano (Eds.)...

  18. [26]

    Jeffrey Pennington, Richard Socher, and Christopher Manning. 2014. GloVe: Global Vectors for Word Representation. InProceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP) , Alessandro Mos- chitti, Bo Pang, and Walter Daelemans (Eds.). As...

  19. [27]

    Alec Radford and Karthik Narasimhan. 2018. Improving Language Understanding by Generative Pre-Training. https://api.semanticscholar.org/CorpusID:49313245

  20. [29]

    Nils Reimers and Iryna Gurevych. 2019. Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks. InProceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Confer- ence on Natural Language Processing (EMNLP-IJ...

  21. [30]

    Victor Sanh, Lysandre Debut, Julien Chaumond, and Thomas Wolf. 2019. Dis- tilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter. ArXiv abs/1910.01108 (2019). https://api.semanticscholar.org/CorpusID:203626972

  22. [31]

    Monika Sester. 2020. Cartographic generalization. Journal of Spatial Information Science 21 (Dec. 2020). https://doi.org/10.5311/josis.2020.21.716

  23. [32]

    Kanae Takahashi, Kouji Yamamoto, Aya Kuchiba, and Tatsuki Koyama. 2022. Confidence interval for micro-averaged F 1 and macro-averaged F 1 scores. Appl. Intell. 52, 5 (March 2022), 4961–4972

  24. [33]

    Michael Tänzer, Sebastian Ruder, and Marek Rei. 2022. Memorisation versus Generalisation in Pre-trained Language Models. InProceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , Smaranda Muresan, Preslav Nakov, and Ali...

  25. [34]

    Gemma Team. 2024. Gemma: Open Models Based on Gemini Research and Technology. arXiv:2403.08295 [cs.CL] https://arxiv.org/abs/2403.08295

  26. [35]

    Hugo Touvron, Thibaut Lavril, Gautier Izacard, Xavier Martinet, Marie-Anne Lachaux, Timothée Lacroix, Baptiste Rozière, Naman Goyal, Eric Hambro, Faisal Azhar, Aurelien Rodriguez, Armand Joulin, Edouard Grave, and Guil- laume Lample. 2023. LLaMA: Open and Efficient Foundation ...

  27. [36]

    Yiqi Wu, Xiaodan Hu, Ziming Fu, Siling Zhou, and Jiangong Li. 2024. GPT-4o: Visual perception performance of multimodal large language models in piglet activity understanding. arXiv:2406.09781 [cs.CV] https://arxiv.org/abs/2406. 09781

  28. [37]

    Wei Zhang, Hao Wei, Bunyamin Sisman, Xin Luna Dong, Christos Faloutsos, and Davd Page. 2020. AutoBlock: A Hands-off Blocking Framework for Entity Match- ing. In Proceedings of the 13th International Conference on Web Search and Data Mining (Houston, TX, USA) (WSDM ’20). Associ...

  29. [38]

    Liu Zhuang, Lin Wayne, Shi Ya, and Zhao Jun. 2021. A Robustly Optimized BERT Pre-training Approach with Post-training. In Proceedings of the 20th Chinese National Conference on Computational Linguistics , Sheng Li, Maosong Sun, Yang Liu, Hua Wu, Kang Liu, Wanxiang Che, Shizhu ...

  30. [2019]

    In 2019 5th International Conference on Web Research (ICWR)

    DeepBlock: A Novel Blocking Approach for Entity Resolution using Deep Learning. In 2019 5th International Conference on Web Research (ICWR) . 41–44. https://doi.org/10.1109/ICWR.2019.8765267

  31. [2020]

    Deep entity matching with pre-trained language models. Proc. VLDB Endow. 14, 1 (sep 2020), 50–60. https://doi.org/10.14778/3421424.3421431

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.