Circuit explained: How does a transformer perform compositional generalization

• Cheng Tang,

  Roles Conceptualization, Formal analysis, Investigation, Methodology, Software, Visualization, Writing – original draft, Writing – review & editing

  Affiliation Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States of America

  ⨯
• Brenden Lake,

  Roles Validation, Writing – review & editing

  Affiliation Departments of Computer Science and Psychology, Princeton University, Princeton, New Jersey, United States of America

  ⨯
• Mehrdad Jazayeri

  Roles Supervision, Writing – review & editing

  * E-mail: mjaz@mit.edu

  Affiliations Department of Brain and Cognitive Sciences, Massachusetts Institute of Technology, Cambridge, Massachusetts, United States of America, Howard Hughes Medical Institute, Cambridge, Massachusetts, United States of America

  https://orcid.org/0000-0002-9764-6961

  ⨯