@inproceedings{valenzuela-etal-2026-holmes,
title = "Holmes-Graph: A Framework for Probing the Linguistic Competence of Graph-Based Text Embeddings",
author = "Valenzuela, Andric Valdez and
Baez Santamaria, Selene and
Gomez Adorno, Helena",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.2172/",
doi = "10.18653/2026.findings-emnlp.2172",
pages = "46482--46493",
ISBN = "979-8-89176-462-0",
abstract = "We present Holmes-Graph, a framework to evaluate the linguistic competences of graph-based text embeddings generated by training graph neural networks (GNNs) on graph representations of text. The framework extends the Holmes benchmark, which previously covered only sparse, static, and transformer-based representations. The framework is flexible to be used with any graph representation and GNN architecture. Once weights are learned on a downstream task, frozen embeddings are probed across five linguistic competences. When applied to AI-generated text detection, Integrated Syntactic Graphs (ISG) combined with Graph Neural Networks (GNN) achieve an aggregate Holmes score of 0.662. This matches or exceeds DeBERTa-base on syntax, morphology, semantics, and reasoning, but not for discourse. The results confirm that graph-based text embeddings encode non-trivial linguistic knowledge and motivate their inclusion in future probing benchmarks."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="valenzuela-etal-2026-holmes">
<titleInfo>
<title>Holmes-Graph: A Framework for Probing the Linguistic Competence of Graph-Based Text Embeddings</title>
</titleInfo>
<name type="personal">
<namePart type="given">Andric</namePart>
<namePart type="given">Valdez</namePart>
<namePart type="family">Valenzuela</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Selene</namePart>
<namePart type="family">Baez Santamaria</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Helena</namePart>
<namePart type="family">Gomez Adorno</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>We present Holmes-Graph, a framework to evaluate the linguistic competences of graph-based text embeddings generated by training graph neural networks (GNNs) on graph representations of text. The framework extends the Holmes benchmark, which previously covered only sparse, static, and transformer-based representations. The framework is flexible to be used with any graph representation and GNN architecture. Once weights are learned on a downstream task, frozen embeddings are probed across five linguistic competences. When applied to AI-generated text detection, Integrated Syntactic Graphs (ISG) combined with Graph Neural Networks (GNN) achieve an aggregate Holmes score of 0.662. This matches or exceeds DeBERTa-base on syntax, morphology, semantics, and reasoning, but not for discourse. The results confirm that graph-based text embeddings encode non-trivial linguistic knowledge and motivate their inclusion in future probing benchmarks.</abstract>
<identifier type="citekey">valenzuela-etal-2026-holmes</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.2172</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2172/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>46482</start>
<end>46493</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Holmes-Graph: A Framework for Probing the Linguistic Competence of Graph-Based Text Embeddings
%A Valenzuela, Andric Valdez
%A Baez Santamaria, Selene
%A Gomez Adorno, Helena
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F valenzuela-etal-2026-holmes
%X We present Holmes-Graph, a framework to evaluate the linguistic competences of graph-based text embeddings generated by training graph neural networks (GNNs) on graph representations of text. The framework extends the Holmes benchmark, which previously covered only sparse, static, and transformer-based representations. The framework is flexible to be used with any graph representation and GNN architecture. Once weights are learned on a downstream task, frozen embeddings are probed across five linguistic competences. When applied to AI-generated text detection, Integrated Syntactic Graphs (ISG) combined with Graph Neural Networks (GNN) achieve an aggregate Holmes score of 0.662. This matches or exceeds DeBERTa-base on syntax, morphology, semantics, and reasoning, but not for discourse. The results confirm that graph-based text embeddings encode non-trivial linguistic knowledge and motivate their inclusion in future probing benchmarks.
%R 10.18653/2026.findings-emnlp.2172
%U https://aclanthology.org/2026.findings-emnlp.2172/
%U https://doi.org/10.18653/2026.findings-emnlp.2172
%P 46482-46493
Markdown (Informal)
[Holmes-Graph: A Framework for Probing the Linguistic Competence of Graph-Based Text Embeddings](https://aclanthology.org/2026.findings-emnlp.2172/) (Valenzuela et al., Findings 2026)
ACL