--- title: "Tom Brown — author profile and papers — Elamur Research" description: "Tom Brown: 2 papers on Elamur Research (2020–2022) — affiliations, topics, co-authors and who cites them in our library." canonical: "https://research.elamur.ai/free/en/authors/tom-brown/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/authors/tom-brown/) [home](https://md.research.elamur.ai/free/en/) / [authors](https://md.research.elamur.ai/free/en/authors/) / Tom Brown # Tom Brown Tom Brown's papers here study how large language models absorb their training text, and what that means for privacy and for model quality. The recurring question is what happens to sequences a model has seen, whether it can reproduce them verbatim or is harmed when they recur too often. The methods range from black-box attacks that probe a trained model from the outside to controlled training runs analysed with scaling laws and mechanistic tools. The work sits between [Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/) and [Interpretability](https://md.research.elamur.ai/free/en/topics/interpretability/). The most cited paper, [Extracting Training Data from Large Language Models](https://md.research.elamur.ai/free/en/papers/carlini-2020-extracting-training-data-large-language/) (2020), challenged the assumption that models trained on vast, deduplicated corpora for few epochs keep their training text private. It showed that someone with only query access can recover exact training sequences. The method first samples many plausible generations and then ranks them by comparing the target model's likelihood with that of a reference model, so that passages the target finds suspiciously probable rise to the top. GPT-2 served as the test case. The newer paper, [Scaling Laws and Interpretability of Learning from Repeated Data](https://md.research.elamur.ai/free/en/papers/hernandez-2022-scaling-laws-interpretability-learning-repeated/) (2022), moves from extraction to the effects of repetition itself. It trains transformer models on corpora that are mostly unique but contain a small, heavily repeated subset, and it varies model size, the share of repeated tokens and the number of repeats. Read through both scaling laws and mechanistic interpretability, the results show a marked double descent in how much the repetition hurts. The extraction study was written with [Nicholas Carlini](https://md.research.elamur.ai/free/en/authors/nicholas-carlini/), [Florian Tramer](https://md.research.elamur.ai/free/en/authors/florian-tramer/), Eric Wallace, Colin Raffel, Dawn Song and others. The repeated-data work was done with [Danny Hernandez](https://md.research.elamur.ai/free/en/authors/danny-hernandez/), Ben Mann and Dawn Drain, among others. Tom Brown has 2 papers in our library, in 2020–2022, most of them on [Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/), [Interpretability](https://md.research.elamur.ai/free/en/topics/interpretability/); often with [Adam Roberts](https://md.research.elamur.ai/free/en/authors/adam-roberts/), [Ariel Herbert-Voss](https://md.research.elamur.ai/free/en/authors/ariel-herbert-voss/), [Ben Mann](https://md.research.elamur.ai/free/en/authors/ben-mann/). 2 papers in our library · 2020–2022 No OpenAlex profile matched yet. [📄 2 works in our library](https://md.research.elamur.ai/free/en/authors/tom-brown/#papers) · [🔗 Cited 703 times in our library](https://md.research.elamur.ai/free/en/authors/tom-brown/cited-by/) · [⭐ Paper Stars ranking: #2216](https://md.research.elamur.ai/free/en/paper-stars/page/23/#star-2216) The profile is the library's match of the name; two people with one name are not always told apart. Elsewhere: [Google Scholar](https://scholar.google.com/citations?view_op=search_authors&mauthors=Tom+Brown) · [Semantic Scholar](https://www.semanticscholar.org/search?q=Tom+Brown) · [dblp](https://dblp.org/search?q=Tom+Brown) · [ACL Anthology](https://aclanthology.org/search/?q=Tom+Brown) ## Fields, topics and concepts Of its works in our library, with the number of works on each. 📚 Shelves 4 The field of science this work belongs to (OpenAlex fields).[Arts and Humanities · 2 works](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/) · [Language and Linguistics · 2 works](https://md.research.elamur.ai/free/en/shelves/arts-and-humanities/language-and-linguistics/) · [Artificial Intelligence · 1 work](https://md.research.elamur.ai/free/en/shelves/computer-science/artificial-intelligence/) · [Computer Science · 1 work](https://md.research.elamur.ai/free/en/shelves/computer-science/) 🧭 Topics 2 Research subjects and languages the work is about.[Language models & pretraining · 1 work](https://md.research.elamur.ai/free/en/topics/language-models/) · [Interpretability · 1 work](https://md.research.elamur.ai/free/en/topics/interpretability/) 💡 Concepts 1 Key ideas and methods explained on our concept pages.[Scaling laws · 1 work](https://md.research.elamur.ai/free/en/concepts/scaling-laws/) ## Works by year ### 2020s [**2022** 1 paper Tom Brown's papers of 2022](https://md.research.elamur.ai/free/en/papers/hernandez-2022-scaling-laws-interpretability-learning-repeated/) · [**2020** 1 paper Tom Brown's papers of 2020](https://md.research.elamur.ai/free/en/papers/carlini-2020-extracting-training-data-large-language/) ## Papers The most cited first (OpenAlex and our library). 1. **[Extracting Training Data from Large Language Models](https://md.research.elamur.ai/free/en/papers/carlini-2020-extracting-training-data-large-language/)** *(2020 · 911 citations)* Large language models are often released after training on scraped or private text. Many assumed they could not leak that text because they barely overfit: GPT-2's training loss is only roughly 10% below its test loss. Carlini et al. show this assumption is wrong. 2. **[Scaling Laws and Interpretability of Learning from Repeated Data](https://md.research.elamur.ai/free/en/papers/hernandez-2022-scaling-laws-interpretability-learning-repeated/)** *(2022 · 84 citations)* Large language models often see the same data several times. Sometimes this is done deliberately, to upweight high-quality sources, and sometimes it results from incomplete deduplication. The authors from Anthropic investigate why even a small fraction of repeated data harms models so severely. They train a family of… ## Co-authors here 12 of 16 co-authors with a page, the most papers together first. - [Adam Roberts](https://md.research.elamur.ai/free/en/authors/adam-roberts/) - [Ariel Herbert-Voss](https://md.research.elamur.ai/free/en/authors/ariel-herbert-voss/) - [Ben Mann](https://md.research.elamur.ai/free/en/authors/ben-mann/) - [Colin Raffel](https://md.research.elamur.ai/free/en/authors/colin-raffel/) - [Danny Hernandez](https://md.research.elamur.ai/free/en/authors/danny-hernandez/) - [Dawn Drain](https://md.research.elamur.ai/free/en/authors/dawn-drain/) - [Dawn Song](https://md.research.elamur.ai/free/en/authors/dawn-song/) - [Eric Wallace](https://md.research.elamur.ai/free/en/authors/eric-wallace/) - [Florian Tramer](https://md.research.elamur.ai/free/en/authors/florian-tramer/) - [Katherine Lee](https://md.research.elamur.ai/free/en/authors/katherine-lee/) - [Matthew Jagielski](https://md.research.elamur.ai/free/en/authors/matthew-jagielski/) - [Nicholas Carlini](https://md.research.elamur.ai/free/en/authors/nicholas-carlini/) ## Cited by Papers of our library cite their works 703 times (of the 141,840 papers whose bibliographies we read); [94 citing papers with a page here →](https://md.research.elamur.ai/free/en/authors/tom-brown/cited-by/). Updated 24 October 2026