{"version":1,"type":"story","url":"https://digestai.news/story/researchers-show-llms-can-link-records-and-infer-ethnicity-from-names","json":"https://digestai.news/story/researchers-show-llms-can-link-records-and-infer-ethnicity-from-names.json","markdown":"https://digestai.news/story/researchers-show-llms-can-link-records-and-infer-ethnicity-from-names.md","slug":"researchers-show-llms-can-link-records-and-infer-ethnicity-from-names","headline":"Researchers show LLMs can link records and infer ethnicity from names without training","summary":"A new paper on arXiv demonstrates that pre-trained language models can perform record linkage and demographic inference from names without needing task-specific training data. The author introduces **EnsembleLink**, an algorithm combining pre-trained embeddings with string-based scorers to match records across scripts and formats, achieving performance comparable to or better than prior methods. For ethnicity inference, large language models reached **83.8% accuracy** on U.S. voter files, surpassing a benchmark of **78.0%** from name-supplemented BISG. The study also shows LLMs can infer group membership for religious and caste-based categories globally, reducing bias in geographic racial gap estimates.\n\nThe paper highlights that pre-trained models encode much of the world knowledge needed for these tasks, minimizing the need for labeled data. Open-weight models were tested and found to approach the performance of proprietary LLMs, offering reproducible implementations for researchers.","keyPoints":["EnsembleLink algorithm matches records across scripts/formats with performance equal to or better than prior methods","LLMs achieve **83.8%** accuracy on U.S. voter files for ethnicity inference, surpassing **78.0%** benchmark","Open-weight models replicate proprietary LLM performance, reducing reliance on labeled training data"],"whyItMatters":"This research proves LLMs can automate costly social science tasks—record matching and demographic inference—without task-specific training, lowering barriers for researchers and policymakers.","category":{"slug":"research","name":"Research","url":"https://digestai.news/category/research"},"entities":{"companies":[],"models":["EnsembleLink","BISG"],"people":[]},"firstPublishedAt":"2026-10-15T04:00:00Z","updatedAt":"2026-10-15T04:00:00Z","sourceCount":1,"hasPrimarySource":true,"sources":[{"outlet":"arXiv cs.CL","title":"Pre-Trained Language Models as Zero-Shot Tools for Social Science Research","url":"https://arxiv.org/abs/2610.19048","publishedAt":"2026-10-15T04:00:00Z","type":"primary","primary":true,"lead":true}],"sourceNotes":null,"discussions":[],"thread":null,"cite":{"text":"Digest AI, \"Researchers show LLMs can link records and infer ethnicity from names without training\", 15 October 2026, https://digestai.news/story/researchers-show-llms-can-link-records-and-infer-ethnicity-from-names","publisher":"Digest AI","title":"Researchers show LLMs can link records and infer ethnicity from names without training","datePublished":"2026-10-15T04:00:00Z","url":"https://digestai.news/story/researchers-show-llms-can-link-records-and-infer-ethnicity-from-names"},"generatedBy":"Written by Digest AI's editorial model from the linked sources; the sources are the record.","license":"Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse"}