Researcher

Marten van Schijndel

Marten van Schijndel is a linguistics professor who studies how people and neural language models process language. He earned a linguistics BA at Western Washington University, completed a PhD at Ohio State, and joined Cornell after a Johns Hopkins postdoc. His papers tested what recurrent networks learn about syntax, examined rogue dimensions in transformer similarity measures, and used sparse autoencoders to trace semantic change in French. In 2026, he coauthored CrosSing, a dataset for evaluating language models’ reasoning about gradable adjectives; he is an associate professor of linguistics at Cornell.

Works
78
Citations
1,015
h-index
17

Publication numbers from OpenAlex, as of 27 October 2026.

Timeline

  1. 2009–2010

    A BA at Western Washington

    At Western Washington University, he earns a BA in linguistics with minors in mathematics, computer science, and Spanish. In 2009, he designs and carries out the first worldwide census of Latin scholars; he is named Classical Language Outstanding Graduate for 2009–2010.

  2. 2012–2013

    A COLING paper and NSF fellowship

    In 2012, he coauthors “Accurate Unbounded Dependency Recovery using Generalized Categorial Grammars” with Luan Nguyen and William Schuler; the paper receives the COLING Best Paper Award. In 2013, he wins an NSF Graduate Research Fellowship for a proposal on general memory constraints in language and expects the fellowship to give him more time for long-term research.

  3. 2017

    A PhD on syntactic frequency

    He completes a PhD in linguistics at Ohio State in 2017, with a dissertation on how syntactic frequencies influence human sentence processing.

  4. By 2019

    A postdoc at Johns Hopkins

    He completes a postdoc at Johns Hopkins University and joins Cornell in 2019.

  5. May 2019

    Joining Cornell’s faculty

    In May 2019, he accepts a tenure-track assistant professor appointment in Linguistics at Cornell, which begins in July 2019.

  6. 2019

    Quantity and syntax in language models

    In 2019, he and Aaron Mueller and Tal Linzen test recurrent neural network language models with different network and training-corpus sizes. They find minimal gains from network growth beyond a certain point and diminishing returns from larger training corpora.

  7. 2021

    Rogue dimensions in transformers

    In 2021, he and William Timkey show that a few rogue dimensions dominate similarity measures for contextualized language models. They find that standardization corrects for rogue dimensions and reveals underlying representational quality.

  8. 2025

    Sparse autoencoders in French text

    In 2025, he and five coauthors use k-sparse autoencoders to interpret contextual word embeddings and trace the semantic evolution of French indigène(s) between 1825 and 1950.

  9. July 2026

    CrosSing tests LLM reasoning

    In July 2026, he coauthors CrosSing, a dataset for testing LLM reasoning about scalar relationships in gradable adjectives against a human baseline. The authors evaluate ten high-performing LLMs and report that some outperform humans when prompts contain no extra information. Model performance falls in certain overinformative contexts.

  10. As of October 2026

    Associate Professor at Cornell

    As of October 2026, he is an Associate Professor of Linguistics at Cornell. He studies how humans and neural language models process language, organizes C.Psyd, and co-organizes Cornell’s Natural Language Processing research group.

Career

  1. NowCornell UniversityAssociate Professor of Linguistics
  2. NowCornell UniversityOrganizer, Computational Psycholinguistic Discussions research group (C.Psyd)
  3. NowCornell UniversityCo-organizer, Natural Language Processing research group
  4. From 2019Cornell UniversityFaculty member
  5. Johns Hopkins UniversityPostdoctoral fellow
  6. Cornell UniversityAssistant Professor of Linguistics

Education

  1. 2017Ohio State UniversityPhD
  2. Western Washington UniversityBA in Linguistics

Sources and credits

Photo: We have no published photo of Marten van Schijndel yet. Send us a free-licensed photo.

  1. WWU names 2009-2010 Outstanding Graduates news.wwu.edu. Checked 27 October 2026.
  2. 2013Q3 Reports: COLING 2012 - Admin Wiki aclweb.org. Checked 27 October 2026.
  3. Spotlight: Graduate NSF Awardees | Department of Linguistics linguistics.osu.edu. Checked 27 October 2026.
  4. OhioLINK ETD: van Schijndel, Marten etd.ohiolink.edu. Checked 27 October 2026.
  5. Marten van Schijndel — Central New York Humanities Corridor cnycorridor.net. Checked 27 October 2026.
  6. Marten van Schijndel accepts job at Cornell | Department of Linguistics linguistics.osu.edu. Checked 27 October 2026.
  7. Quantity doesn’t buy quality syntax with neural language models - ACL Anthology aclanthology.org. Checked 27 October 2026.
  8. All Bark and No Bite: Rogue Dimensions in Transformer Language Models Obscure Representational Quality - ACL Anthology aclanthology.org. Checked 27 October 2026.
  9. Disentangling language change: sparse autoencoders quantify the semantic evolution of indigeneity in French - ACL Anthology aclanthology.org. Checked 27 October 2026.
  10. CrosSing: Cross-Scale Reasoning Evaluation on LLMs against Humans - ACL Anthology aclanthology.org. Checked 27 October 2026.
  11. Marten van Schijndel | Department of Linguistics linguistics.cornell.edu. Checked 27 October 2026.