Hendrik Strobelt

Career

Strobelt holds a doctorate in computer science, in visualization, from the University of Konstanz, and held postdoctoral positions at Harvard SEAS and NYU Tandon before joining IBM in 2017. Strobelt is a Senior Research Scientist at IBM Research and the Explainability Lead for the MIT-IBM Watson AI Lab, and is a visiting researcher at MIT CSAIL. IBM Research's profile also notes temporary faculty positions at Carnegie Mellon University and the University of Konstanz.¹

Research

Strobelt's research centers on visualization for, and human collaboration with, AI models, aimed at explainability and intuition; earlier work concerned NLP and generative models, and the methods have been applied in machine learning, biomedicine, and chemistry. Recent co-authored papers listed by IBM Research include work on multi-level explanations for generative language models (ACL 2025) and on interactive counterfactual analysis of large language model behavior (IEEE TVCG 2025).¹

In chemistry, Strobelt co-authored a 2021 IBM Research post on RXNMapper, an open-source atom-mapping tool that learned the grammar of organic chemistry reactions without human supervision, developed by IBM Research Europe, the MIT-IBM Watson AI Lab, and the University of Bern.⁵

Open-source work and tools

Strobelt is first author of LSTMVis (2016), a tool for exploring hidden-state dynamics in recurrent neural networks, and a co-author of GLTR (2019), an open-source tool that uses a language model's own statistics to help people spot generated text. In 2020 Strobelt built a NeurIPS paper visualization tool, built with Benjamin Hoover, that clusters accepted papers by theme using language-model embeddings and UMAP.⁴²³⁶

Scientific service

Strobelt's work has appeared at venues including IEEE VIS, ICLR, ACL, NeurIPS, and ICCV, and in journals such as PNAS and Science Advances. Strobelt has served on organizing committees for IEEE VIS, VISxAI, ICLR, ICML, and NeurIPS.¹

Work and contributions

• GLTR (Giant Language Model Test Room) Open-source code

  Co-author of "GLTR: Statistical Detection and Visualization of Generated Text" (2019); the Apache-2.0 repository describes GLTR as a project by Hendrik Strobelt, Sebastian Gehrmann, and Alexander M. Rush, built as a collaboration of the MIT-IBM Watson AI Lab and HarvardNLP.²³

• LSTMVis Published research

  First author of "LSTMVis: A Tool for Visual Analysis of Hidden State Dynamics in Recurrent Neural Networks" (2016), with Sebastian Gehrmann, Hanspeter Pfister, and Alexander M. Rush.⁴

• Granite Switch 4.1 8B Open weights

  Listed among the authors of the Hugging Face blog post introducing Granite Switch 4.1, which describes the Granite-Switch-4.1-8B model.⁷

Timeline

1. Jun 23, 2016Released LSTMVis, a visual analysis tool for recurrent neural networks, as first author.⁴
2. 2017Joined IBM after postdoctoral positions at Harvard SEAS and NYU Tandon.¹
3. Jun 10, 2019Co-authored GLTR, a tool for statistical detection and visualization of generated text.²
4. 2020Built a visualization tool for finding and presenting NeurIPS papers.⁶
5. Apr 7, 2021Co-authored IBM Research's post on RXNMapper and the grammar of organic chemistry.⁵

Other documented roles

Roles the sources document that are not confirmed as current.

• Postdoctoral position at Harvard SEAS (before joining IBM in 2017), Harvard University (SEAS)¹
• Postdoctoral position at NYU Tandon (before joining IBM in 2017), NYU Tandon School of Engineering¹

Where they work

• IBM 29 people in the wiki

Related people

• Luis LastrasAlso at IBM
• Michael HindAlso at IBM
• Rama AkkirajuAlso at IBM
• Kush R. VarshneyAlso at IBM
• Sriram RaghavanAlso at IBM
• Jacob AndreasLanguage models · Interpretability
• David BauInterpretability · Language models
• Arthur ConmyInterpretability · Language models
• Nick FrosstLanguage models · Interpretability
• Samuel MarksInterpretability · Language models