@inproceedings{henderson-etal-2023-transformers,
title = "Transformers as Graph-to-Graph Models",
author = "Henderson, James and
Mohammadshahi, Alireza and
Coman, Andrei and
Miculicich, Lesly",
editor = "Elazar, Yanai and
Ettinger, Allyson and
Kassner, Nora and
Ruder, Sebastian and
A. Smith, Noah",
booktitle = "Proceedings of the Big Picture Workshop",
month = dec,
year = "2023",
address = "Singapore",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2023.bigpicture-1.8",
doi = "10.18653/v1/2023.bigpicture-1.8",
pages = "93--107",
abstract = "We argue that Transformers are essentially graph-to-graph models, with sequences just being a special case. Attention weights are functionally equivalent to graph edges. Our Graph-to-Graph Transformer architecture makes this ability explicit, by inputting graph edges into the attention weight computations and predicting graph edges with attention-like functions, thereby integrating explicit graphs into the latent graphs learned by pretrained Transformers. Adding iterative graph refinement provides a joint embedding of input, output, and latent graphs, allowing non-autoregressive graph prediction to optimise the complete graph without any bespoke pipeline or decoding strategy. Empirical results show that this architecture achieves state-of-the-art accuracies for modelling a variety of linguistic structures, integrating very effectively with the latent linguistic representations learned by pretraining.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="henderson-etal-2023-transformers">
<titleInfo>
<title>Transformers as Graph-to-Graph Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">James</namePart>
<namePart type="family">Henderson</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Alireza</namePart>
<namePart type="family">Mohammadshahi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andrei</namePart>
<namePart type="family">Coman</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lesly</namePart>
<namePart type="family">Miculicich</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2023-12</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Big Picture Workshop</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yanai</namePart>
<namePart type="family">Elazar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Allyson</namePart>
<namePart type="family">Ettinger</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nora</namePart>
<namePart type="family">Kassner</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sebastian</namePart>
<namePart type="family">Ruder</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Noah</namePart>
<namePart type="family">A. Smith</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Singapore</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>We argue that Transformers are essentially graph-to-graph models, with sequences just being a special case. Attention weights are functionally equivalent to graph edges. Our Graph-to-Graph Transformer architecture makes this ability explicit, by inputting graph edges into the attention weight computations and predicting graph edges with attention-like functions, thereby integrating explicit graphs into the latent graphs learned by pretrained Transformers. Adding iterative graph refinement provides a joint embedding of input, output, and latent graphs, allowing non-autoregressive graph prediction to optimise the complete graph without any bespoke pipeline or decoding strategy. Empirical results show that this architecture achieves state-of-the-art accuracies for modelling a variety of linguistic structures, integrating very effectively with the latent linguistic representations learned by pretraining.</abstract>
<identifier type="citekey">henderson-etal-2023-transformers</identifier>
<identifier type="doi">10.18653/v1/2023.bigpicture-1.8</identifier>
<location>
<url>https://aclanthology.org/2023.bigpicture-1.8</url>
</location>
<part>
<date>2023-12</date>
<extent unit="page">
<start>93</start>
<end>107</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Transformers as Graph-to-Graph Models
%A Henderson, James
%A Mohammadshahi, Alireza
%A Coman, Andrei
%A Miculicich, Lesly
%Y Elazar, Yanai
%Y Ettinger, Allyson
%Y Kassner, Nora
%Y Ruder, Sebastian
%Y A. Smith, Noah
%S Proceedings of the Big Picture Workshop
%D 2023
%8 December
%I Association for Computational Linguistics
%C Singapore
%F henderson-etal-2023-transformers
%X We argue that Transformers are essentially graph-to-graph models, with sequences just being a special case. Attention weights are functionally equivalent to graph edges. Our Graph-to-Graph Transformer architecture makes this ability explicit, by inputting graph edges into the attention weight computations and predicting graph edges with attention-like functions, thereby integrating explicit graphs into the latent graphs learned by pretrained Transformers. Adding iterative graph refinement provides a joint embedding of input, output, and latent graphs, allowing non-autoregressive graph prediction to optimise the complete graph without any bespoke pipeline or decoding strategy. Empirical results show that this architecture achieves state-of-the-art accuracies for modelling a variety of linguistic structures, integrating very effectively with the latent linguistic representations learned by pretraining.
%R 10.18653/v1/2023.bigpicture-1.8
%U https://aclanthology.org/2023.bigpicture-1.8
%U https://doi.org/10.18653/v1/2023.bigpicture-1.8
%P 93-107
Markdown (Informal)
[Transformers as Graph-to-Graph Models](https://aclanthology.org/2023.bigpicture-1.8) (Henderson et al., BigPicture 2023)
ACL
- James Henderson, Alireza Mohammadshahi, Andrei Coman, and Lesly Miculicich. 2023. Transformers as Graph-to-Graph Models. In Proceedings of the Big Picture Workshop, pages 93–107, Singapore. Association for Computational Linguistics.