PeopleofSuperintelligence

Researcher

Amanda Bertsch

Amanda Bertsch researches language models, long-context reasoning, and inference algorithms. She entered the University of Arizona as a Flinn Scholar in 2017, studied computer science and mathematics, and published research on Wikipedia puffery and minimum Bayes risk decoding. Her long-context learning work won NAACL's 2025 SAC Award for Language Modeling, and she designed and co-taught a language-model inference course before coauthoring OLMo 3. In 2026, two papers she coauthored were accepted to COLM; as of October 2026, she is a final-year PhD candidate at Carnegie Mellon and a student researcher at AI2.

Works
33
Citations
178
h-index
7

Publication numbers from OpenAlex, as of 18 October 2026.

Timeline

  1. 2017

    A Flinn Scholar at Arizona

    In 2017, she joins the University of Arizona's College of Engineering as a Flinn Scholar after graduating as valedictorian from Tri-City College Prep in Prescott.

  2. 2021

    A Wikipedia puffery paper

    In 2021, she is a senior double-majoring in computer science and mathematics at the University of Arizona. She and Steven Bethard publish a W-NUT paper that uses citation features in a RoBERTa model to identify puffery on Wikipedia.

  3. 2023

    Minimum Bayes risk decoding

    In 2023, she and three coauthors show how several modern language-generation methods can be expressed as minimum Bayes risk decoding and report theoretical and empirical results for its variants.

  4. 2025

    Long-context in-context learning

    In 2025, she and six coauthors study in-context learning with thousands of demonstrations and show that performance rises for many datasets with large label spaces. The work wins NAACL's 2025 SAC Award for Language Modeling.

  5. 2025

    Teaching language-model inference

    In 2025, she designs and co-teaches a new course, Inference Algorithms for Language Modeling.

  6. 2025

    Coauthoring OLMo 3

    In 2025, she coauthors OLMo 3, a family of fully open 7B and 32B language models that targets long-context reasoning, function calling, coding, instruction following, general chat, and knowledge recall.

  7. 2026

    Two papers reach COLM

    In 2026, COLM accepts two papers she coauthors: Oolong, a benchmark for long-context reasoning and aggregation, and Cracks in the Foundation, a study of transformer choices that affect long-context extension. Oolong asks models to analyze and aggregate individual text chunks; Cracks shows that combinations of architectural choices can limit long-context performance.

  8. As of October 2026

    A PhD at Carnegie Mellon

    As of October 2026, she is a final-year PhD candidate in Carnegie Mellon's Language Technologies Institute and a student researcher at the Allen Institute for AI.

Career

  1. Since 2025Allen Institute for AI (Ai2)Student researcher

Education

  1. From 2017University of Arizona
  2. Carnegie Mellon UniversityPhD

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

Most-cited works

  1. Bridging the Gap: A Survey on Integrating (Human) Feedback for Natural Language Generation2023 · Transactions of the Association for Computational Linguistics · 44 citations
  2. Unlimiformer: Long-Range Transformers with Unlimited Length Input2023 · neural information processing systems · 29 citations
  3. SummQA at MEDIQA-Chat 2023: In-Context Learning with GPT-4 for Medical Summarization2023 · Clinical Natural Language Processing Workshop (ClinicalNLP) · 18 citations
  4. LLMs as Workers in Human-Computational Algorithms? Replicating Crowdsourcing Pipelines with LLMs2025 · Extended Abstracts of the CHI Conference on Human Factors in Computing Systems (CHI EA) · 17 citations
  5. In-Context Learning with Long-Context Models: An In-Depth Exploration2025 · Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technologies (NAACL) · 11 citations

From OpenAlex, as of 18 October 2026.

  1. The company share of AI researchers’ top papers rose from 3% to 32%Includes Allen Institute for AI (Ai2). We estimate where leading AI researchers worked when their most-cited papers appeared. The company share rose from 3% in 1975–1979 to 32% in 2020–2024.
  2. Google has the highest combined h-index among AI organizationsIncludes Allen Institute for AI (Ai2). Our estimates of combined and median h-index at AI organizations, using current affiliations and OpenAlex records.
  3. 44% of AI career moves go from one university to anotherIncludes Allen Institute for AI (Ai2). We estimate how 12,774 AI career moves link universities, companies, labs and government research centers, with net flows and the most common paths.
  4. More people shaping AI studied at Massachusetts Institute of Technology than at any other institutionIncludes Carnegie Mellon University. Our estimates of where the people shaping AI studied and where their careers led, from education records in cited biographies.

Colleagues

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

Sources and credits

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

  1. Flinn Scholars lay foundation for engineering careers news.engineering.arizona.edu. Checked 18 October 2026.
  2. Awards 2020 - 2021 | Computer Science cs.arizona.edu. Checked 18 October 2026.
  3. Detection of Puffery on the English Wikipedia aclanthology.org. Checked 18 October 2026.
  4. It's MBR All the Way Down: Modern Generation Techniques Through the Lens of Minimum Bayes Risk arxiv.org. Checked 18 October 2026.
  5. In-Context Learning with Long-Context Models: An In-Depth Exploration aclanthology.org. Checked 18 October 2026.
  6. Amanda Bertsch cs.cmu.edu. Checked 18 October 2026.
  7. Olmo 3 arxiv.org. Checked 18 October 2026.
  8. Oolong: Evaluating Long Context Reasoning and Aggregation Capabilities arxiv.org. Checked 18 October 2026.
  9. COLM 2026 Accepted Papers colm.eventhosts.cc. Checked 18 October 2026.
  10. Cracks in the Foundation: Seemingly Minor Architectural Choices Impact Long Context Extension arxiv.org. Checked 18 October 2026.