PeopleofSuperintelligence

Researcher

Shane Arora

Shane Arora is an AI researcher whose work spans mathematical analysis, long-form question answering, and open language models. A former Honours student at the University of Sydney's School of Mathematics and Statistics, he coauthored papers on Fréchet derivatives in 2019 and nonlinear eigenvalue problems in 2020. In 2023, he studied how retrieval-augmented language models use evidence in long-form answers; at Ai2, he coauthored OLMo and OLMoE in 2024, CaLMQA and a training-efficiency study in 2025, and OLMo 2 and Olmo 3. In 2025, he is a researcher at Ai2.

Works
11
Citations
100
h-index
4

Publication numbers from OpenAlex, as of 22 October 2026.

Timeline

  1. 2019

    Fréchet derivatives on Banach spaces

    A former Honours student at the University of Sydney's School of Mathematics and Statistics, he coauthors a 2019 paper with Hazel Browne and Daniel Daners on Fréchet derivatives. The paper reduces many differentiability questions on Banach spaces to questions of continuity.

  2. 2020

    Nonlinear eigenvalue branches

    In 2020, he studies solution branches of nonlinear eigenvalue problems on restricted domains. For points of odd multiplicity, the component of solutions through the bifurcation point is unbounded, accumulates at the boundary, or contains an even number of odd-multiplicity points.

  3. 2023

    Retrieval for long-form answers

    In 2023, he coauthors two controlled studies of retrieval-augmented language models for long-form question answering, varying the evidence documents or the models. The study tracks how generated answers attribute claims to retrieved evidence documents.

  4. 2024

    OLMo and OLMoE

    In 2024, as an Ai2 researcher, he coauthors OLMo, an open language model built for scientific study and released with open training data and training and evaluation code. He also coauthors OLMoE, a fully open mixture-of-experts model with 7 billion parameters and 1 billion active per input token.

  5. 2025

    The CaLMQA dataset

    In 2025, he coauthors CaLMQA, a dataset of 51.7K culturally specific questions across 23 languages. It defines culturally specific questions as concepts unique to one or a few cultures or questions with answers that vary by cultural or regional context.

  6. 2025

    Critical batch size

    In 2025, he coauthors a study that measures critical batch size during OLMo training. Batch-size warmup trains OLMo 1B to slightly better loss with 43% fewer gradient steps.

  7. 2025

    OLMo 2 and Olmo 3

    In 2025, he coauthors OLMo 2, the next generation of fully open language models. In December 2025, he coauthors Olmo 3, a fully open model family at 7B and 32B parameter scales.

  8. 2025

    Researcher at Ai2

    In 2025, he is a researcher at the Allen Institute for Artificial Intelligence (Ai2).

Career

  1. 2024–2025Allen Institute for AI (Ai2)Researcher
  2. 2024–2025University of Massachusetts Amherst

Education

  1. University of SydneyHonours student

From the biography and OpenAlex affiliations (the years of papers that list the organization), so some years are approximate.

Most-cited works

  1. OLMo: Accelerating the Science of Language Models2024 · Annual Meeting of the Association for Computational Linguistics (ACL) · 69 citationsWith Crystal Nam, Dustin Schwenk, Hamish Ivison, Ian Magnusson, Jacob Morrison and Kyle Lo
  2. OLMoE: Open Mixture-of-Experts Language Models2024 · 8 citationsWith Ali Farhadi, Dustin Schwenk, Jacob Morrison, Kyle Lo, Nathan Lambert and Niklas Muennighoff
  3. Olmo 32025 · 5 citationsWith Akari Asai, Allyson Ettinger, Aman Rangapur, Amanda Bertsch, Ashish Sabharwal and Bailey Kuehl
  4. CaLMQA: Exploring culturally specific long-form question answering across 23 languages2025 · Annual Meeting of the Association for Computational Linguistics (ACL) · 4 citations
  5. 2 OLMo 2 Furious2024 · 2 citationsWith Aman Rangapur, Christopher Clark, Crystal Nam, Dustin Schwenk, Faeze Brahman and Hamish Ivison

From OpenAlex, as of 21 October 2026.

  1. The company share of AI researchers’ top papers was 31% in 2020–2024 and 8% in 1970–1974Includes Allen Institute for AI (Ai2). We estimate where the people shaping AI worked when their most-cited papers appeared: the company share was 31% in 2020–2024 and 8% in 1970–1974, and Google led all organizations in 2020–2024.
  2. Companies supply the most prior-work hires at 7 of the 9 frontier AI labsIncludes Allen Institute for AI (Ai2). Our estimates of the organizations where 550 people worked directly before joining 9 frontier AI labs across 620 career transitions, comparing the main feeder organizations per lab.
  3. Universities keep AI researchers for a median of 6 years, compared with 2 years at frontier labsIncludes University of Massachusetts Amherst. Across 15,811 dated work stays (6,529 people), median role tenure runs from 6 years at universities to 2 years at frontier labs. We compare all five organization types, starting decades from the 1970s to the 2020s, and the organizations that keep people longest.
  4. 72% of AI researchers with top works connect into a single 2,592-person co-author groupIncludes Allen Institute for AI (Ai2). We map the co-author network across 3,593 people shaping AI and 14,977 most-cited papers: 2,592 (72%) form one connected group with a median distance of 5 steps, led in 2-step reach by Yoshua Bengio (570) and in shortest-path bridging by Yoshua Bengio (6.7%).

Colleagues

People with dated positions at the same organization in the same years.

Sources and credits

Photo: We have no published photo of Shane Arora yet. Send us a free-licensed photo.

  1. S_Arora - USyd Maths & Stats mail.maths.usyd.edu.au. Checked 21 October 2026.
  2. An alternative approach to Fréchet derivatives arxiv.org. Checked 21 October 2026.
  3. Solution branches of nonlinear eigenvalue problems on restricted domains arxiv.org. Checked 21 October 2026.
  4. Understanding Retrieval Augmentation for Long-Form Question Answering arxiv.org. Checked 21 October 2026.
  5. A way with words and trillions of tokens: Allen School researchers recognized at ACL 2024 for expanding the capabilities of language models news.cs.washington.edu. Checked 21 October 2026.
  6. OLMo: Accelerating the Science of Language Models arxiv.org. Checked 21 October 2026.
  7. OLMoE: Open Mixture-of-Experts Language Models arxiv.org. Checked 21 October 2026.
  8. CaLMQA: Exploring culturally specific long-form question answering across 23 languages aclanthology.org. Checked 21 October 2026.
  9. Critical Batch Size Revisited: A Simple Empirical Approach to Large-Batch Language Model Training proceedings.neurips.cc. Checked 21 October 2026.
  10. 2 OLMo 2 Furious arxiv.org. Checked 21 October 2026.
  11. Olmo 3 arxiv.org. Checked 21 October 2026.
  12. Apple Workshop on Natural Language and Interactive Systems 2025 machinelearning.apple.com. Checked 21 October 2026.