aiqbench analysis
This paper proposes EnsembleLink, a record-linkage method that uses pre-trained language models to match entities across datasets without labeled training data, and reports accuracy comparable to labeling-heavy approaches on several benchmarks. Its practical appeal is that it runs on open-source models locally. Readers should note that the arXiv listing marks this version as superseded by arXiv:2610.19048, so the merged version is the one to check for current results.
EnsembleLink: Accurate Record Linkage Without Training Data
arXiv:2601.21138v3 Announce Type: replace Abstract: Record linkage, the process of matching records that refer to the same entity across datasets, is essential to empirical social science but remains methodologically underdeveloped. Researchers treat it as a preprocessing step, applying ad hoc rules without quantifying the uncertainty that linkage errors introduce into downstream analyses. Existing methods either achieve low accuracy or require substantial labeled training data. I present EnsembleLink, a method that achieves high accuracy without any training labels. EnsembleLink leverages pre-trained language models that have learned semantic relationships (e.g., that "South Ozone Park" is a neighborhood in "New York City" or that "Lutte ouvriere" refers to the Trotskyist "Workers' Struggle" party) from large text corpora. On benchmarks spanning city names, person names, organizations, multilingual political parties, and bibliographic records, EnsembleLink matches or exceeds methods requiring extensive labeling. The method runs locally on open-source models, requiring no external API calls, and completes typical linkage tasks in minutes.
Abstract:Record linkage, the process of matching records that refer to the same entity across datasets, is essential to empirical social science but remains methodologically underdeveloped. Researchers treat it as a preprocessing step, applying ad hoc rules without quantifying the uncertainty that linkage errors introduce into downstream analyses. Existing methods either achieve low accuracy or require substantial labeled training data. I present EnsembleLink, a method that achieves high accuracy without any training labels. EnsembleLink leverages pre-trained language models that have learned semantic relationships (e.g., that "South Ozone Park" is a neighborhood in "New York City" or that "Lutte ouvriere" refers to the Trotskyist "Workers' Struggle" party) from large text corpora. On benchmarks spanning city names, person names, organizations, multilingual political parties, and bibliographic records, EnsembleLink matches or exceeds methods requiring extensive labeling. The method runs locally on open-source models, requiring no external API calls, and completes typical linkage tasks in minutes.
This paper has been superseded by arXiv:2610.19048 (merged from arXiv:2601.21132 and arXiv:2601.21138)
Computation and Language (cs.CL)
arXiv:2601.21138 [cs.CL]
(or arXiv:2601.21138v3 [cs.CL] for this version)
https://doi.org/10.48550/arXiv.2601.21138 Focus to learn more arXiv-issued DOI via DataCite
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 255 words
Related coverage
Pre-Trained Language Models as Zero-Shot Tools for Social Science Research
arXiv · Oct 15, 2026
Multi-Vector (Late Interaction) Embedding Models with Sentence Transformers
Hugging Face · Aug 18, 2026
Expenditure Horizon: Measuring Optimization Ability, with an Application to NanoGPT
METR · Jul 21, 2026
Links #3: 2026/06 Part 1
LessWrong · Jun 15, 2026