ACL 2022long92 citations

Making Transformers Solve Compositional Tasks

Santiago Ontanon, Joshua Ainslie, Zachary Fisher, Vaclav Cvicek

Abstract

Several studies have reported the inability of Transformer models to generalize compositionally, a key type of generalization in many NLP tasks such as semantic parsing. In this paper we explore the design space of Transformer models showing that the inductive biases given to the model by several design decisions significantly impact compositional generalization. We identified Transformer configurations that generalize compositionally significantly better than previously reported in the literature in many compositional tasks. We achieve state-of-the-art results in a semantic parsing compositional generalization benchmark (COGS), and a string edit operation composition benchmark (PCFG).

BibTeX
@inproceedings{ontanon-etal-2022-making,
    title = "Making Transformers Solve Compositional Tasks",
    author = "Ontanon, Santiago  and
      Ainslie, Joshua  and
      Fisher, Zachary  and
      Cvicek, Vaclav",
    editor = "Muresan, Smaranda  and
      Nakov, Preslav  and
      Villavicencio, Aline",
    booktitle = "Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = may,
    year = "2022",
    address = "Dublin, Ireland",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2022.acl-long.251/",
    doi = "10.18653/v1/2022.acl-long.251",
    pages = "3591--3607"
}
Making Transformers Solve Compositional Tasks · ACL 2022