@inproceedings{kulumba-etal-2026-halvest,
title = "{HAL}vest-Contrastive: Retrieval-like Authorship Attribution with Patch-level Late Interaction",
author = "Kulumba, Francis and
Antoun, Wissam and
Vimont, Guillaume and
Romary, Laurent and
Cafiero, Florian",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.2660/",
pages = "60552--60571",
ISBN = "979-8-89176-459-0",
abstract = "Authorship attribution is the process of identifying the writer of a disputed text by analyzing distinctive stylistic patterns. By comparing the disputed text against a baseline of known authorship, the task amounts to assessing how strongly the two texts are likely to share an author. However, topical confounding makes the task deceptively easy: two authors covering the same topic may look more alike than one author covering two topics. Scholarly prose offers a natural solution. Academic writers produce multiple papers on related but distinct topics while maintaining consistent stylistic habits. Thus, we introduce HALvest-Contrastive, an English benchmark that draws same-author passages from distinct papers within a disciplinary field to reduce topical overlap, while using different-author passages from the same field to discourage naive keyword matching. The benchmark{'}s design removes topical shortcuts: a strong lexical baseline fails under these constraints. As a by-product of our benchmark construction, we release the 17B-token multilingual corpus used to produce it. We then revisit how to score authorship: while standard systems compress each document into a single vector, we retain the sequence of token-level representations produced by a language model and compare them with late interaction. We also study patch-level late interaction, which groups neighboring tokens into patches before interaction. Matching at the token level increases Recall@20 by a factor of three to five. The optimal interaction granularity, however, depends on sequence length, with no single granularity performing best across all settings."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="kulumba-etal-2026-halvest">
<titleInfo>
<title>HALvest-Contrastive: Retrieval-like Authorship Attribution with Patch-level Late Interaction</title>
</titleInfo>
<name type="personal">
<namePart type="given">Francis</namePart>
<namePart type="family">Kulumba</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wissam</namePart>
<namePart type="family">Antoun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Guillaume</namePart>
<namePart type="family">Vimont</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Laurent</namePart>
<namePart type="family">Romary</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Florian</namePart>
<namePart type="family">Cafiero</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Authorship attribution is the process of identifying the writer of a disputed text by analyzing distinctive stylistic patterns. By comparing the disputed text against a baseline of known authorship, the task amounts to assessing how strongly the two texts are likely to share an author. However, topical confounding makes the task deceptively easy: two authors covering the same topic may look more alike than one author covering two topics. Scholarly prose offers a natural solution. Academic writers produce multiple papers on related but distinct topics while maintaining consistent stylistic habits. Thus, we introduce HALvest-Contrastive, an English benchmark that draws same-author passages from distinct papers within a disciplinary field to reduce topical overlap, while using different-author passages from the same field to discourage naive keyword matching. The benchmark’s design removes topical shortcuts: a strong lexical baseline fails under these constraints. As a by-product of our benchmark construction, we release the 17B-token multilingual corpus used to produce it. We then revisit how to score authorship: while standard systems compress each document into a single vector, we retain the sequence of token-level representations produced by a language model and compare them with late interaction. We also study patch-level late interaction, which groups neighboring tokens into patches before interaction. Matching at the token level increases Recall@20 by a factor of three to five. The optimal interaction granularity, however, depends on sequence length, with no single granularity performing best across all settings.</abstract>
<identifier type="citekey">kulumba-etal-2026-halvest</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2660/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>60552</start>
<end>60571</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T HALvest-Contrastive: Retrieval-like Authorship Attribution with Patch-level Late Interaction
%A Kulumba, Francis
%A Antoun, Wissam
%A Vimont, Guillaume
%A Romary, Laurent
%A Cafiero, Florian
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F kulumba-etal-2026-halvest
%X Authorship attribution is the process of identifying the writer of a disputed text by analyzing distinctive stylistic patterns. By comparing the disputed text against a baseline of known authorship, the task amounts to assessing how strongly the two texts are likely to share an author. However, topical confounding makes the task deceptively easy: two authors covering the same topic may look more alike than one author covering two topics. Scholarly prose offers a natural solution. Academic writers produce multiple papers on related but distinct topics while maintaining consistent stylistic habits. Thus, we introduce HALvest-Contrastive, an English benchmark that draws same-author passages from distinct papers within a disciplinary field to reduce topical overlap, while using different-author passages from the same field to discourage naive keyword matching. The benchmark’s design removes topical shortcuts: a strong lexical baseline fails under these constraints. As a by-product of our benchmark construction, we release the 17B-token multilingual corpus used to produce it. We then revisit how to score authorship: while standard systems compress each document into a single vector, we retain the sequence of token-level representations produced by a language model and compare them with late interaction. We also study patch-level late interaction, which groups neighboring tokens into patches before interaction. Matching at the token level increases Recall@20 by a factor of three to five. The optimal interaction granularity, however, depends on sequence length, with no single granularity performing best across all settings.
%U https://aclanthology.org/2026.emnlp-main.2660/
%P 60552-60571
Markdown (Informal)
[HALvest-Contrastive: Retrieval-like Authorship Attribution with Patch-level Late Interaction](https://aclanthology.org/2026.emnlp-main.2660/) (Kulumba et al., EMNLP 2026)
ACL