--- title: "Sanjeev Arora — author profile and papers" description: "Sanjeev Arora, Princeton University: 3 papers on Elamur Research (2016–2024), h-index 68 — affiliations, topics, co-authors and who cites them in our…" canonical: "https://research.elamur.ai/free/en/authors/sanjeev-arora/" --- [Elamur Research](https://md.research.elamur.ai/free/en/) · [this page in HTML](https://research.elamur.ai/free/en/authors/sanjeev-arora/) [home](https://md.research.elamur.ai/free/en/) / [authors](https://md.research.elamur.ai/free/en/authors/) / Sanjeev Arora # Sanjeev Arora Sanjeev Arora's papers here study the mathematics of how machine learning models represent language. The earlier work builds generative, latent-variable models of text. These models explain why word vectors learned from co-occurrence statistics behave as they do, and the papers back the explanations with proofs rather than experiments alone. The themes fall under [lexical semantics and word embeddings](https://md.research.elamur.ai/free/en/topics/lexical-semantics/) and [word sense disambiguation](https://md.research.elamur.ai/free/en/topics/word-sense-disambiguation/). The more recent work asks how the choice of training data shapes large language models. The most cited of these papers is [A Latent Variable Model Approach to PMI-based Word Embeddings](https://md.research.elamur.ai/free/en/papers/arora-2016-latent-variable-model-approach-pmi/) (2016). It treats text as produced by a hidden discourse vector that drifts slowly as a random walk, with each word emitted according to its inner product with that vector. The model accounts for two observations: low-dimensional vectors closely approximate PMI matrices, and analogies can be solved with vector arithmetic. [Linear Algebraic Structure of Word Senses, with Applications to Polysemy](https://md.research.elamur.ai/free/en/papers/arora-2018-linear-algebraic-structure-word-senses/) (2018) carries this framework over to ambiguous words. It argues that the embedding of a word with several meanings is a weighted, nonnegative combination of vectors for its separate senses. Using a Gaussian version of the walk model, it proves that a word's vector is a linear transformation of the averaged embeddings of its contexts. The newest work, [LESS: Selecting Influential Data for Targeted Instruction Tuning](https://md.research.elamur.ai/free/en/papers/xia-2024-less-selecting-influential-data-targeted/) (2024), moves into [language models and pretraining](https://md.research.elamur.ai/free/en/topics/language-models/). It addresses a setting where a model needs only a narrow set of skills, and where training on a broad instruction mixture can even weaken those skills. Given a few examples of the desired capability, the method ranks items in a large instruction pool by a gradient-based influence estimate. That estimate is adapted to the Adam optimiser and compared by cosine similarity. Arora is based at [Princeton University](https://md.research.elamur.ai/free/en/universities/princeton-university/). Earlier bylines include the Institute for Advanced Study and the [University of California, Berkeley](https://md.research.elamur.ai/free/en/universities/university-of-california-berkeley/). Co-authors on these papers include [Andrej Risteski](https://md.research.elamur.ai/free/en/authors/andrej-risteski/), [Tengyu Ma](https://md.research.elamur.ai/free/en/authors/tengyu-ma/), [Yingyu Liang](https://md.research.elamur.ai/free/en/authors/yingyu-liang/) and [Yuanzhi Li](https://md.research.elamur.ai/free/en/authors/yuanzhi-li/), as well as [Mengzhou Xia](https://md.research.elamur.ai/free/en/authors/mengzhou-xia/) and [Danqi Chen](https://md.research.elamur.ai/free/en/authors/danqi-chen/). Sanjeev Arora has 3 papers in our library, in 2016–2024, most of them on [Word sense disambiguation](https://md.research.elamur.ai/free/en/topics/word-sense-disambiguation/), [Language models & pretraining](https://md.research.elamur.ai/free/en/topics/language-models/), [Lexical semantics & word embeddings](https://md.research.elamur.ai/free/en/topics/lexical-semantics/); often with [Andrej Risteski](https://md.research.elamur.ai/free/en/authors/andrej-risteski/), [Tengyu Ma](https://md.research.elamur.ai/free/en/authors/tengyu-ma/), [Yingyu Liang](https://md.research.elamur.ai/free/en/authors/yingyu-liang/). 3 papers in our library · 2016–2024 [OpenAlex](https://openalex.org/A5103209777) · [ORCID](https://orcid.org/0000-0002-8636-9970) · [Wikidata](https://www.wikidata.org/wiki/Q92664) · [**h-index 68**](https://openalex.org/authors/A5103209777) · [348 works](https://openalex.org/works?filter=author.id:A5103209777) · [22,918 citations](https://openalex.org/works?filter=author.id:A5103209777&sort=cited_by_count:desc) OpenAlex, 2026-10-14 [📄 3 works in our library](https://md.research.elamur.ai/free/en/authors/sanjeev-arora/#papers) · [🔗 705 citations in our library](https://md.research.elamur.ai/free/en/authors/sanjeev-arora/cited-by/) · [7 more in our library without a page yet](#also) · [⭐ Paper Star rank #2319](https://md.research.elamur.ai/free/en/paper-stars/page/24/#star-2319) 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=Sanjeev+Arora) · [Semantic Scholar](https://www.semanticscholar.org/search?q=Sanjeev+Arora) · [dblp](https://dblp.org/search?q=Sanjeev+Arora) · [ACL Anthology](https://aclanthology.org/search/?q=Sanjeev+Arora) ## Affiliations [🎓**Princeton University** Princeton · United States 1995–2026 · 213 papersEmployment*current*](https://md.research.elamur.ai/free/en/universities/princeton-university/) · [🔬**Institute for Advanced Study** Princeton · United States 2019–2020 · 10 papers](https://ror.org/00f809463) · [🎓**University of California, Berkeley** Berkeley · United States 1992–2002 · 5 papersEducation](https://md.research.elamur.ai/free/en/universities/university-of-california-berkeley/) · [🎓**Yale University** New Haven · United States 2005–2021 · 2 papers](https://md.research.elamur.ai/free/en/universities/yale-university/) · [🎓**Massachusetts Institute of Technology** Cambridge · United States 1990 · 1 paperEducation](https://md.research.elamur.ai/free/en/universities/massachusetts-institute-of-technology/) From the bylines of their papers (OpenAlex, 2026-10-14), ORCID and Wikidata; institutions on papers are matched by machine. 2 more institutions on single bylines are not listed. ## Topics [Word sense disambiguation (1)](https://md.research.elamur.ai/free/en/topics/word-sense-disambiguation/) · [Language models & pretraining (1)](https://md.research.elamur.ai/free/en/topics/language-models/) · [Lexical semantics & word embeddings (1)](https://md.research.elamur.ai/free/en/topics/lexical-semantics/) ## Works by year ### 2020s [**2024** 1 paper](https://md.research.elamur.ai/free/en/papers/xia-2024-less-selecting-influential-data-targeted/) ### 2010s [**2018** 1 paper](https://md.research.elamur.ai/free/en/papers/arora-2018-linear-algebraic-structure-word-senses/) · [**2016** 1 paper](https://md.research.elamur.ai/free/en/papers/arora-2016-latent-variable-model-approach-pmi/) ## Papers The most cited first (OpenAlex and our library). 1. **[A Latent Variable Model Approach to PMI-based Word Embeddings](https://md.research.elamur.ai/free/en/papers/arora-2016-latent-variable-model-approach-pmi/)** *(2016 · 365 citations)* Embeddings derived from co-occurrence statistics, such as PMI matrices, word2vec and GloVe, work well in practice. 2. **[Linear Algebraic Structure of Word Senses, with Applications to Polysemy](https://md.research.elamur.ai/free/en/papers/arora-2018-linear-algebraic-structure-word-senses/)** *(2018 · 247 citations)* The paper asks what a single word vector encodes when the word has several meanings. The authors argue that standard embeddings such as word2vec and GloVe store a polysemous word as a weighted, nonnegative sum of vectors for its individual senses. They support this "Linearity Assertion" with theory. They modify the… 3. **[LESS: Selecting Influential Data for Targeted Instruction Tuning](https://md.research.elamur.ai/free/en/papers/xia-2024-less-selecting-influential-data-targeted/)** *(2024 · 170 citations)* Instruction tuning usually merges many heterogeneous datasets, but real deployments often need only a narrow set of skills. ## Also in our library 7 more works of theirs are in our library without a page here yet; they will appear as we publish them. ## Co-authors here 7 co-authors with a page, the most papers together first. - [Andrej Risteski](https://md.research.elamur.ai/free/en/authors/andrej-risteski/) ×2 - [Tengyu Ma](https://md.research.elamur.ai/free/en/authors/tengyu-ma/) ×2 - [Yingyu Liang](https://md.research.elamur.ai/free/en/authors/yingyu-liang/) ×2 - [Yuanzhi Li](https://md.research.elamur.ai/free/en/authors/yuanzhi-li/) ×2 - [Danqi Chen](https://md.research.elamur.ai/free/en/authors/danqi-chen/) - [Mengzhou Xia](https://md.research.elamur.ai/free/en/authors/mengzhou-xia/) - [Suchin Gururangan](https://md.research.elamur.ai/free/en/authors/suchin-gururangan/) ## Cited by Papers of our library cite their works 705 times (of the 140,884 papers whose bibliographies we read); [24 citing papers with a page here →](https://md.research.elamur.ai/free/en/authors/sanjeev-arora/cited-by/).