Leveraging Language Models for Automated Patient Record Linkage

Iris Zachary; Lovedeep Gondara; Mohammad Beheshti

REVIEW 5 major objections 5 minor 31 references

Leveraging Language Models for Automated Patient Record Linkage

T0 review · 5 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

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

Pith's one-line read Fine-tuned Mistral-7B errs on only 6 of 52,917 patient pairs

desk verdict Useful applied benchmark for LLM-based patient matching, but the near-perfect accuracy is measured against labels produced by the same probabilistic linkage tool whose fields the models consume, so read the headline numbers as agreement with Match*Pro, not ground truth. read the letter →

arxiv 2504.15261 v1 pith:RHBTXJNP submitted 2025-04-21 cs.AI cs.LG

Mohammad Beheshti , Lovedeep Gondara , Iris Zachary This is my paper · ORCID

classification cs.AIcs.LG

keywords patientrecordlinkageblockingmatchingfine-tunedLLMszero-shotcancerregistryRoBERTaMistral

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

This paper asks whether language models can take over the two steps of patient record linkage: blocking, which cuts the pool of possible record pairs, and matching, which decides whether a pair refers to the same person. Using real cancer-registry records, the authors report that a fine-tuned 7-billion-parameter model made only 6 wrong match decisions among 52,917 test pairs, and that a fine-tuned RoBERTa embedding model cut the candidate-pair pool by 92% while missing just 3 of 2,356 true matches. The authors conclude that fine-tuned LLMs are a scalable alternative to hand-built matching rules, while acknowledging that embedding-based blocking is less robust to typos than a hybrid rule-based and probabilistic approach. If these results generalize, registries could automate most of the clerical review that currently keeps linkage accurate.

What carries the argument

The mechanism that carries the argument is pair serialization plus transformer fine-tuning. Each record is converted to a token sequence with attribute markers, two records are joined with separator tokens, and the model is trained to output a match or non-match decision; for blocking, a single record is serialized into one string, embedded by a fine-tuned RoBERTa using mean pooling, and candidate pairs are found by vector nearest-neighbor search with a cosine-similarity cutoff. This pipeline supplies both headline numbers: the matching errors come from the fine-tuned generative models on the serialized pairs, and the blocking reduction comes from the embedding distance threshold.

What would settle it

Take a random sample of the 52,917 test pairs, reclassify them using an identifier outside the linkage fields, such as a medical record number, a tissue block number, or a direct patient query, and recompute each model's false positives and false negatives against that independent standard; if the independent labels disagree with the probabilistic-linkage labels in even a few dozen cases, the near-perfect scores are not measuring true linkage accuracy.

Watch

Extended reading notes

Core claim

The central claim is that language models can perform patient record matching at near-human error rates and can also do useful blocking. On held-out 2021 data, the fine-tuned Mistral-7B matcher produced 0 false positives and 6 false negatives, for a total of 6 errors and an F1 score of 0.9993; the best zero-shot model, Mistral-Small-24B, produced 55 total errors. For blocking, a fine-tuned RoBERTa encoder with K=10 nearest neighbors and a cosine threshold of 0.75 generated 4,250 candidate pairs instead of the 52,917 in the baseline, a 92% reduction, while missing 3 true matches. The paper also claims that a hybrid of rule-based blocking plus a probabilistic similarity threshold would have reduced pairs even further, by 95%, while maintaining 100% recall, and that chain-of-thought reasoning models are impractical for large-scale linkage because of inference time.

Load-bearing premise

The load-bearing premise is that the human-reviewed labels produced by the probabilistic linkage tool are a reliable ground truth; those labels are built from the same name, birth date, sex, SSN, and address fields the models read, so any error in the labels would make the reported near-perfect error counts an underestimate of true linkage error.

Editorial extensions

If this is right

Fine-tuned LLM matching could reduce manual review workload in registries: with 6 errors in 52,917 pairs, only a tiny fraction of decisions would need human adjudication.
For surveillance applications that require every true match to be found, the 3 matches lost at the 0.75 threshold make embedding blocking risky; the paper's own data suggest a hybrid rule-based threshold would preserve 100% recall.
A small fine-tuned model (Mistral-7B) outperforms much larger zero-shot models, so domain fine-tuning matters more than raw parameter count for this task.
Zero-shot performance is still strong, with the best model making 55 errors at an F1 score of 0.994, meaning LLMs can be used without any training when labeled data are scarce.
Reasoning models such as DeepSeek-R1 are not viable for tens of thousands of pairs: 26 hours versus 30 minutes on a 2,736-pair subset.

Reading between the lines

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

Because the labels were generated by a probabilistic linker that scores the same fields the models consume, a portion of the reported accuracy may reflect learning the linker's scoring tendencies rather than an independent truth; a registry with a true unique patient identifier would give a cleaner test.
Injecting controlled single-character typos into names and birth dates would directly test the paper's subword-tokenizer explanation for blocking failures, and could predict how much real-world dirty data would hurt.
Routing only the ambiguous 0.85-0.95 similarity band to a fine-tuned LLM, and letting a deterministic rule handle the rest, could combine the 95% blocking reduction with near-perfect recall.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Desk Editor's Note

read the letter

What you should know: this is one of the first papers to test fine-tuned LLMs on real patient-level record linkage from a cancer registry, with actual PHI. The headline numbers are strong—Mistral-7B fine-tuned makes 6 errors on ~53k test pairs, RoBERTa blocking cuts 92% of candidate pairs. But those numbers are measured against labels created by Match*Pro, a probabilistic linkage tool that scores the same identifiers (name, DOB, SSN, sex) the models read. So the near-perfect matching may reflect learning to agree with Match*Pro rather than resolving true identity. That caveat doesn't kill the paper, but it changes what the results mean.

What's genuinely new: applying these models to a real registry linkage task with heavily missing SSN/Address (97%/81%) is a needed step beyond benchmark entity matching. The blocking result is the most valuable part: embedding-based similarity struggles with minor typos and is both less accurate and less efficient than a hybrid rule-based + probabilistic approach. That's a concrete, non-obvious negative finding. The paper is also honest about limitations—it acknowledges the difficulty of thresholds, the reliance on the Overall Similarity Score, and the impractical compute of DeepSeek-R1.

Soft spots: (1) The ground truth is not independent. Manual review helps, but with SSN and Address mostly absent, both Match*Pro and human annotators lean on names and birth date. The models could be approximating the annotators' conventions rather than some external truth. This makes the 6-error claim a statement about agreement with the registry's current process, not verified identity. (2) Blocking hyperparameters (K and cosine threshold) are selected on the test set, so the 92% reduction is optimistic. (3) The DeepSeek comparison is on a post hoc subset (similarity 0.65–1.0), making that cost/performance contrast less clean. (4) No code or data—expected with PHI, but it blocks independent replication.

None of these are fatal. The paper is a solid applied contribution with a useful negative result, and the authors are transparent about what they did. I'd accept it for peer review, but the reviewers should push for either an independent gold standard subset or at least an error analysis on the cases where Match*Pro and the models disagree. The paper promises a "scalable solution" for automation—that claim is conditional on the labels being trustworthy.

Recommendation: send it to peer review, and treat the headline error counts as agreement-with-Match*Pro rather than ground-truth accuracy. The blocking negative is the keeper.

Referee Report

5 major / 5 minor

Summary. The paper investigates using language models for two stages of patient record linkage on real-world Missouri Cancer Registry data: blocking (candidate-pair generation) and matching (binary match/non-match classification). The authors fine-tune RoBERTa to produce sentence embeddings for blocking, and fine-tune or zero-shot prompt several open-source LLMs (plus a RoBERTa classifier) for matching, using a probabilistic-linkage output (Match*Pro) with manual review as the ground-truth labels. The headline results are a 92% reduction in candidate pairs for blocking (K=10, cosine threshold 0.75) and only 6 incorrect predictions for fine-tuned Mistral-7B on 52,917 test pairs, with Mistral-Small-24B the best zero-shot model at 55 errors. The discussion concludes that language models are promising for automation but that a hybrid rule-based/probabilistic approach remains preferable for blocking, and that DeepSeek-R1 is impractical due to computational cost.

Significance. If the results were validated against an independent gold standard, the paper would be a valuable feasibility study: it demonstrates that open-source LLMs can reproduce registry linkage decisions with very low error and that embedding-based blocking can shrink the comparison space on real PHI data. The study has genuine strengths: it uses real patient data from a cancer registry, compares a wide range of open models, and is transparent about data missingness (SSN and Address missing in 97% and 81% of dataset B records, respectively). However, the current evidence does not support the strongest claims: the labels are produced by the same probabilistic linkage tool whose identifiers the models consume, and several key thresholds are selected on the test set. The reported performance is therefore best interpreted as agreement with Match*Pro's scoring conventions rather than verified identity resolution. The paper also contains inconsistent F1 calculations that need correction.

major comments (5)

[Data source and preprocessing; Table 6] The ground-truth labels for dataset AB are generated by Match*Pro's Fellegi–Sunter configuration over exactly the identifiers (First Name, Middle Name, Last Name, Sex, Birth Date, SSN, Address) that are later serialized into the matching prompts. Although the pairs are manually reviewed, SSN and Address are missing in 97% and 81% of dataset B records, so both the annotators and the models rely almost entirely on names and birth date/month. The reported FP+FN counts in Table 6 (e.g., 6 for fine-tuned Mistral-7B) therefore measure agreement with the registry's existing linkage labels rather than verified patient identity; this should be stated as a central limitation, and the headline claim of 'only 6 incorrect predictions' should be reworded or supported with an external gold standard.
[Experiment 1: blocking model; Figure 3] The blocking model is fine-tuned using labels derived from Match*Pro's Overall Similarity Score, and the optimal K and cosine similarity threshold are chosen by searching directly on the test set (Figure 3). Consequently, the 92% candidate-pair reduction and near-perfect recall are optimistic and may not generalize; the hyperparameters should be selected on a validation split and the final recall/pair-reduction numbers reported on a held-out test set. Additionally, because the training labels come from the same probabilistic linkage process, the evaluation does not establish that embedding-based blocking outperforms or complements rule-based blocking on an independent criterion.
[Discussion: 'Language models for blocking'] The claim that a hybrid rule-based and probabilistic approach outperforms the language-model blocker is based on applying an Overall Similarity Score threshold of 0.65 that is chosen post hoc from the test data (Figure 2), yielding 2,736 candidate pairs with 100% recall. Using the same test set to select the threshold and to compute the recall overstates the hybrid approach and makes the comparison unfair. A fair comparison would pre-specify the threshold on training/validation data or use a separate tuning set.
[Matching model performance (DeepSeek subset)] DeepSeek-R1-70B is evaluated only on the 2,736 test pairs selected by the post hoc criterion 0.65 < Overall Similarity Score < 1.0, so its comparison with Mistral-Small-24B and Llama-3.3-70B on that subset inherits the same test-set-selection bias. The conclusion that reasoning models are impractical for record linkage should be restricted to this subset and to the chosen threshold; the 26-hour versus 30-minute runtime comparison is meaningful only for this filtered set.
[Table 6] The F1 scores in Table 6 do not match the reported FP and FN columns under the standard binary F1 formula. For example, fine-tuned Mistral-7B (FP=0, FN=6) should have F1 = 2×2350/(2×2350+6) = 0.9987, not 0.9993, and zero-shot Mistral-7B (FP=2450, FN=4) should be 0.657, not 0.816. Because the text uses 'nearly perfect F1 scores' to justify focusing on FP+FN, these values must be recomputed or the F1 definition stated.

minor comments (5)

[Table 3 caption] The caption 'List of Language Models Used in Experiment 1' should read 'Experiment 2' because the table lists matching models, not blocking models.
[Classification model setup] The reference to 'the training arguments listed in Table 1' should be to Table 4 (Training Arguments Used for Fine-Tuning the Models for the Matching Task), since Table 1 describes the Match*Pro linkage configuration.
[Abstract] The abstract states that fine-tuned Mistral-7B made 'only 6 incorrect predictions' without noting that this is a matching-stage result on the pre-linked test set; please add a qualifier such as 'among the pre-linked candidate pairs' to avoid implying an end-to-end error rate.
[Table 3 and Table 5] The model name is given inconsistently as DeepSeek-R1-70B in the text and as DeepSeek-R1-Distill-Llama-70B in Tables 3 and 5; please unify the terminology.
[Discussion] The phrase 'a hybrid rule-based and probabilistic approach' is vague; it should specify that this refers to Match*Pro with an Overall Similarity Score threshold, so the comparison is explicit.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the language models are trained and evaluated on externally constructed Match*Pro/manual-review labels, not on labels derived from the models' own outputs.

full rationale

The paper's derivation chain is self-contained: Match*Pro produces Match/Non-Match labels (with manual review) from the same identifiers that are later serialized into model inputs. This is supervised learning against a noisy gold standard, not a derivation that reduces to its inputs. At inference, the models receive only serialized identifiers (or Ditto-style attribute tokens); they do not receive the Overall Similarity Score or Match*Pro's output. The blocking experiment trains embeddings with cosine similarity loss using labels derived from the Overall Similarity Score and then evaluates candidate retrieval against the separately human-reviewed AB binary labels; while the label source is not an error-free gold standard, the evaluation target is not the model's own training objective. The acknowledged missing SSN/Address data weakens the gold standard's authority but does not make the reported predictions definitionally equal to the inputs. No self-citation chain, uniqueness theorem, ansatz smuggling, or renaming of known results appears in the paper.

Assumptions & free parameters 3 free parameters · 4 assumptions · 0 invented entities

The central evaluation depends on Match*Pro-derived labels as ground truth, on the sufficiency of the available identifiers, on the temporal stability from 2022 to 2021, and on cosine similarity as a blocking signal. The blocking parameters K and threshold were selected on the test set, adding two fit-to-test parameters.

free parameters (3)

K (number of nearest neighbors) = 10
Selected by evaluating different K values on the test set; increasing beyond 10 provided no recall gain.
Cosine similarity threshold (blocking) = 0.75
Selected as the optimal balance between recall and candidate pair count on the test data; lower thresholds give more pairs, higher thresholds hurt recall.
Similarity score cutoff for DeepSeek subset = 0.65
Used to restrict DeepSeek-R1 evaluation to edge cases after inspecting test distribution; not a pre-specified design.

assumptions (4)

domain assumption Fellegi-Sunter probabilistic linkage with Match*Pro produces valid ground truth labels for record linkage.
The whole evaluation assumes the Match*Pro labels (with manual review) are correct; if the probabilistic model is biased, all reported error counts are relative to a biased target.
domain assumption The set of identifiers (name, DOB, SSN, sex, address) is sufficient to distinguish patients.
No unique patient identifier exists in dataset B, and SSN/address are missing in 97%/81% of records, so true matches are unknown; the linkage assumes these fields are discriminating.
domain assumption Temporal transfer from 2022 training to 2021 test is valid.
The models are fine-tuned on 2022 data and tested on 2021 data; this assumes the record linkage task is stationary across years, which may not hold if data quality changes.
ad hoc to paper Cosine similarity of embedding vectors is a meaningful measure of record similarity for blocking.
The paper adopts cosine similarity without evidence that it correlates with true matches better than existing phonetic blocking for this data; indeed the results show it is suboptimal.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Leveraging Language Models for Automated Patient Record Linkage." pith.science (2026). https://pith.science/paper/RHBTXJNP

@misc{pith2026250415261,
  author       = {Pith},
  title        = {Pith review of: Leveraging Language Models for Automated Patient Record Linkage},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/RHBTXJNP}},
  note         = {Machine review of arXiv:2504.15261}
}

read the original abstract

Objective: Healthcare data fragmentation presents a major challenge for linking patient data, necessitating robust record linkage to integrate patient records from diverse sources. This study investigates the feasibility of leveraging language models for automated patient record linkage, focusing on two key tasks: blocking and matching. Materials and Methods: We utilized real-world healthcare data from the Missouri Cancer Registry and Research Center, linking patient records from two independent sources using probabilistic linkage as a baseline. A transformer-based model, RoBERTa, was fine-tuned for blocking using sentence embeddings. For matching, several language models were experimented under fine-tuned and zero-shot settings, assessing their performance against ground truth labels. Results: The fine-tuned blocking model achieved a 92% reduction in the number of candidate pairs while maintaining near-perfect recall. In the matching task, fine-tuned Mistral-7B achieved the best performance with only 6 incorrect predictions. Among zero-shot models, Mistral-Small-24B performed best, with a total of 55 incorrect predictions. Discussion: Fine-tuned language models achieved strong performance in patient record blocking and matching with minimal errors. However, they remain less accurate and efficient than a hybrid rule-based and probabilistic approach for blocking. Additionally, reasoning models like DeepSeek-R1 are impractical for large-scale record linkage due to high computational costs. Conclusion: This study highlights the potential of language models for automating patient record linkage, offering improved efficiency by eliminating the manual efforts required to perform patient record linkage. Overall, language models offer a scalable solution that can enhance data integration, reduce manual effort, and support disease surveillance and research.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 18 canonical work pages

[1]

Big data for health

Andreu-Perez J, … CP … journal of biomedical, 2015 undefined. Big data for health. ieeexplore.ieee.orgJ Andreu-Perez, CCY Poon, RD Merrifield, STC Wong, GZ YangIEEE journal of biomedical and health informatics, 2015•ieeexplore.ieee.org [Internet]. [cited 2025 Jan 27]; Available from: https://ieeexplore.ieee.org/abstract/document/7154395/

work page arXiv 2015
[2]

Cancer Statistics | CDC [Internet]

Incidence Data Sources | U.S. Cancer Statistics | CDC [Internet]. [cited 2025 Jan 27]. Available from: https://www.cdc.gov/united-states-cancer-statistics/technical- notes/incidence-data-sources.html

work page 2025
[3]

Record linkage: Current practice and future directions

Gu L, Baxter R, Vickers D, and CRCM, 2003 undefined. Record linkage: Current practice and future directions. CiteseerL Gu, R Baxter, D Vickers, C RainsfordCSIRO Mathematical and Information Sciences Technical Report, 2003•Citeseer [Internet]. [cited 2025 Jan 27]; Available from: https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=a2c4dec86a96a ...

work page 2003
[4]

A Theory for Record Linkage

Fellegi IP , Sunter AB. A Theory for Record Linkage. J Am Stat Assoc. 1969;64(328):1183–210

work page 1969
[5]

Probabilistic record linkage

Sayers A, Ben-Shlomo Y , Blom AW, Steele F . Probabilistic record linkage. Int J Epidemiol [Internet]. 2015 Jun 1 [cited 2025 Jan 27];45(3):954. Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC5005943/

work page 2015
[6]

Demystifying probabilistic linkage: Common myths and misconceptions

Doidge JC, Harron K. Demystifying probabilistic linkage: Common myths and misconceptions. Int J Popul Data Sci [Internet]. 2018 Jan 10 [cited 2025 Jan 27];3(1):410. Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC6281162/

work page 2018
[7]

Bert: Pre-training of deep bidirectional transformers for language understanding

Devlin J, Chang M, Lee K, arXiv KT arXiv preprint, 2018 undefined. Bert: Pre-training of deep bidirectional transformers for language understanding. arxiv.org [Internet]. [cited 2025 Jan 28]; Available from: https://arxiv.org/abs/1810.04805?amp=1

arXiv 2018
[8]

Attention is all you need

Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez AN, et al. Attention is all you need. Adv Neural Inf Process Syst. 2017;2017-December:5999–6009

work page 2017

Show all 31 references

[9]

Deep entity matching with pre-trained language models

Li Y , Li J, Suhara Y , Doan A, arXiv:2004.00584 WT arXiv preprint, 2020 undefined. Deep entity matching with pre-trained language models. arxiv.org [Internet]. 2021 Sep 1 [cited 2025 Jan 28];14(1):50–60. Available from: https://arxiv.org/abs/2004.00584

2004 arXiv
[10]

Fine-tuning Large Language Models for Entity Matching

Steiner A, Peeters R, Bizer C. Fine-tuning Large Language Models for Entity Matching. 2024 Sep 12 [cited 2025 Jan 28]; Available from: http://arxiv.org/abs/2409.08185

2024 arXiv
[11]

Entity Matching using Large Language Models

Peeters R, Steiner A, Bizer C. Entity Matching using Large Language Models. 2023 Oct 17 [cited 2025 Jan 28]; Available from: http://arxiv.org/abs/2310.11244

2023 arXiv
[12]

A Comparison of Blocking Methods for Record Linkage

Steorts RC, Ventura SL, Sadinle M, Fienberg SE. A Comparison of Blocking Methods for Record Linkage. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) [Internet]. 2014 [cited 2025 Mar 4];8744:25...

2014 doi
[13]

[cited 2025 Mar 2]

Match*Pro Software - SEER Registrars [Internet]. [cited 2025 Mar 2]. Available from: https://seer.cancer.gov/tools/matchpro/

2025
[14]

Sudowoodo: Contrastive self-supervised learning for multi-purpose data integration and preparation

Wang R, Li Y , International JW 2023 I 39th, 2023 undefined. Sudowoodo: Contrastive self-supervised learning for multi-purpose data integration and preparation. ieeexplore.ieee.orgR Wang, Y Li, J Wang2023 IEEE 39th International Conference on Data Engineering (ICDE), 2023•ieee...

2023
[15]

Duplicate Detection with GenAI

Ormesher I. Duplicate Detection with GenAI. 2024 Jun 17 [cited 2025 Mar 4]; Available from: https://arxiv.org/abs/2406.15483v1

2024 arXiv
[16]

LinkTransformer: A Unified Package for Record Linkage with Transformer Language Models

Arora A, Dell M. LinkTransformer: A Unified Package for Record Linkage with Transformer Language Models. Proceedings of the Annual Meeting of the Association for Computational Linguistics [Internet]. 2024 [cited 2025 Mar 4];3:221–31. Available from: https://linktransformer.github.io/

2024
[17]

Heterogeneous Entity Matching with Complex Attribute Associations using BERT and Neural Networks

Wang S, Lu J. Heterogeneous Entity Matching with Complex Attribute Associations using BERT and Neural Networks. 2023 Sep 19 [cited 2025 Mar 4]; Available from: http://arxiv.org/abs/2309.11046

2023 arXiv
[18]

Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks

arXiv:1908.10084 NR arXiv preprint, 2019 undefined. Sentence-BERT: Sentence Embeddings using Siamese BERT-Networks. fq.pkwyx.com [Internet]. [cited 2025 Jan 31];3982–92. Available from: https://fq.pkwyx.com/default/https/aclanthology.org/D19-1410.pdf

1908 arXiv
[19]

RoBERTa: A Robustly Optimized BERT Pretraining Approach

Liu Y , Ott M, Goyal N, Du J, Joshi M, Chen D, et al. RoBERTa: A Robustly Optimized BERT Pretraining Approach. 2019 Jul 26 [cited 2025 Jan 31]; Available from: https://arxiv.org/abs/1907.11692v1

2019 arXiv
[20]

Explaining Text Similarity in Transformer Models

Vasileiou A, Eberle O. Explaining Text Similarity in Transformer Models. 2024 May 10 [cited 2025 Feb 17]; Available from: http://arxiv.org/abs/2405.06604

2024 arXiv
[21]

The faiss library

Douze M, Guzhva A, Deng C, Johnson J, Szilvasy G, Mazaré PE, et al. The faiss library. arXiv preprint arXiv:240108281. 2024

2024
[22]

LLaMA: Open and Efficient Foundation Language Models

Touvron H, Lavril T, Izacard G, Martinet X, Lachaux MA, Lacroix T, et al. LLaMA: Open and Efficient Foundation Language Models. 2023 Feb 27 [cited 2025 Mar 10]; Available from: http://arxiv.org/abs/2302.13971

2023 arXiv
[23]

Mistral 7B

Jiang AQ, Sablayrolles A, Mensch A, Bamford C, Chaplot DS, Casas D de las, et al. Mistral 7B. 2023 Oct 10 [cited 2025 Mar 10]; Available from: https://arxiv.org/abs/2310.06825v1

2023 arXiv
[25]

[cited 2025 Mar 2]

RoBERTa [Internet]. [cited 2025 Mar 2]. Available from: https://huggingface.co/docs/transformers/en/model_doc/roberta#transformers.Rob ertaForSequenceClassification

2025
[26]

Lora: Low-rank adaptation of large language models

Hu E, Shen Y , Wallis P , Allen-Zhu Z, Li Y , Wang S, et al. Lora: Low-rank adaptation of large language models. arxiv.orgEJ Hu, Y Shen, P Wallis, Z Allen-Zhu, Y Li, S Wang, L Wang, W ChenICLR, 2022•arxiv.org [Internet]. [cited 2025 Mar 2]; Available from: https://arxiv.org/pd...

2022 arXiv
[27]

http://github.com/unslothai/unsloth

Daniel Han MH and U team. http://github.com/unslothai/unsloth. 2023. Unsloth

2023
[28]

Language models are few-shot learners

Brown TB, Mann B, Ryder N, Subbiah M, Kaplan J, Dhariwal P , et al. Language models are few-shot learners. proceedings.neurips.cc [Internet]. [cited 2025 Feb 14]; Available from: https://proceedings.neurips.cc/paper/2020/hash/1457c0d6bfcb4967418bfb8ac142f 64a-Abstract.html

2025
[29]

Scaling Laws for Neural Language Models

Kaplan J, McCandlish S, Henighan T, Brown TB, Chess B, Child R, et al. Scaling Laws for Neural Language Models. 2020 Jan 22 [cited 2025 Mar 6]; Available from: http://arxiv.org/abs/2001.08361

2020 arXiv
[30]

Emergent abilities of large language models

Wei J, Tay Y , Bommasani R, Raffel C, Zoph B, Borgeaud S, et al. Emergent abilities of large language models. arxiv.org [Internet]. [cited 2025 Mar 6]; Available from: https://arxiv.org/abs/2206.07682

2025 arXiv
[31]

Large language models are zero-shot reasoners

Kojima T, Shane Gu S, Reid Google Research M, Matsuo Y , Iwasawa Y . Large language models are zero-shot reasoners. proceedings.neurips.ccT Kojima, SS Gu, M Reid, Y Matsuo, Y IwasawaAdvances in neural information processing systems, 2022•proceedings.neurips.cc [Internet]. [cit...

2022
[32]

Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning

Guo D, Yang D, Zhang H, Song J, … RZ arXiv preprint arXiv, 2025 undefined. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arxiv.org [Internet]. 2025 [cited 2025 Mar 6]; Available from: https://arxiv.org/abs/2501.12948

2025 arXiv

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.

[1] [1]

Big data for health

Andreu-Perez J, … CP … journal of biomedical, 2015 undefined. Big data for health. ieeexplore.ieee.orgJ Andreu-Perez, CCY Poon, RD Merrifield, STC Wong, GZ YangIEEE journal of biomedical and health informatics, 2015•ieeexplore.ieee.org [Internet]. [cited 2025 Jan 27]; Available from: https://ieeexplore.ieee.org/abstract/document/7154395/

work page arXiv 2015

[2] [2]

Cancer Statistics | CDC [Internet]

Incidence Data Sources | U.S. Cancer Statistics | CDC [Internet]. [cited 2025 Jan 27]. Available from: https://www.cdc.gov/united-states-cancer-statistics/technical- notes/incidence-data-sources.html

work page 2025

[3] [3]

Record linkage: Current practice and future directions

Gu L, Baxter R, Vickers D, and CRCM, 2003 undefined. Record linkage: Current practice and future directions. CiteseerL Gu, R Baxter, D Vickers, C RainsfordCSIRO Mathematical and Information Sciences Technical Report, 2003•Citeseer [Internet]. [cited 2025 Jan 27]; Available from: https://citeseerx.ist.psu.edu/document?repid=rep1&type=pdf&doi=a2c4dec86a96a ...

work page 2003

[4] [4]

A Theory for Record Linkage

Fellegi IP , Sunter AB. A Theory for Record Linkage. J Am Stat Assoc. 1969;64(328):1183–210

work page 1969

[5] [5]

Probabilistic record linkage

Sayers A, Ben-Shlomo Y , Blom AW, Steele F . Probabilistic record linkage. Int J Epidemiol [Internet]. 2015 Jun 1 [cited 2025 Jan 27];45(3):954. Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC5005943/

work page 2015

[6] [6]

Demystifying probabilistic linkage: Common myths and misconceptions

Doidge JC, Harron K. Demystifying probabilistic linkage: Common myths and misconceptions. Int J Popul Data Sci [Internet]. 2018 Jan 10 [cited 2025 Jan 27];3(1):410. Available from: https://pmc.ncbi.nlm.nih.gov/articles/PMC6281162/

work page 2018

[7] [7]

Bert: Pre-training of deep bidirectional transformers for language understanding

Devlin J, Chang M, Lee K, arXiv KT arXiv preprint, 2018 undefined. Bert: Pre-training of deep bidirectional transformers for language understanding. arxiv.org [Internet]. [cited 2025 Jan 28]; Available from: https://arxiv.org/abs/1810.04805?amp=1

arXiv 2018

[8] [8]

Attention is all you need

Vaswani A, Shazeer N, Parmar N, Uszkoreit J, Jones L, Gomez AN, et al. Attention is all you need. Adv Neural Inf Process Syst. 2017;2017-December:5999–6009

work page 2017