@inproceedings{dat-etal-2026-eventkd,
title = "{E}vent{KD}: Event-Aware Knowledge Distillation for Generative Event Extraction",
author = "Dat, Hoang Thanh and
Dao, Quoc Phong and
Hai, Nam Le and
Dernoncourt, Franck and
Van, Linh Ngo and
Le, Trung and
Nguyen, Thien Huu",
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.154/",
pages = "3240--3253",
ISBN = "979-8-89176-462-0",
abstract = "Event extraction aims to identify triggers in text, classify them into predefined event types, and recognize associated arguments such as participants and locations. Large language model (LLM)-based generative approaches provide a unified paradigm for joint trigger identification, classification, and argument extraction; however, their computational cost necessitates knowledge distillation (KD) into smaller student models. Existing KD methods operate at the token level, treating each output position independently and thus failing to concentrate the distillation signal on event-relevant spans or to capture the relational structure among event components. We propose EventKD, a two-level distillation framework tailored for event extraction. At the token level, a distributional divergence loss incorporates event-aware span importance weighting to prioritize tokens corresponding to triggers, arguments, and type labels. At the span level, a Relational-Aware (RA) loss preserves pairwise geometric relationships among event-aware spans across Transformer layers. Experiments on four benchmarks (ACE05, MAVEN, GENEVA, and RAMS) demonstrate that EventKD consistently outperforms existing distillation methods on both trigger identification and argument extraction."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="dat-etal-2026-eventkd">
<titleInfo>
<title>EventKD: Event-Aware Knowledge Distillation for Generative Event Extraction</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hoang</namePart>
<namePart type="given">Thanh</namePart>
<namePart type="family">Dat</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Quoc</namePart>
<namePart type="given">Phong</namePart>
<namePart type="family">Dao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nam</namePart>
<namePart type="given">Le</namePart>
<namePart type="family">Hai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Franck</namePart>
<namePart type="family">Dernoncourt</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Linh</namePart>
<namePart type="given">Ngo</namePart>
<namePart type="family">Van</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Trung</namePart>
<namePart type="family">Le</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Thien</namePart>
<namePart type="given">Huu</namePart>
<namePart type="family">Nguyen</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>Event extraction aims to identify triggers in text, classify them into predefined event types, and recognize associated arguments such as participants and locations. Large language model (LLM)-based generative approaches provide a unified paradigm for joint trigger identification, classification, and argument extraction; however, their computational cost necessitates knowledge distillation (KD) into smaller student models. Existing KD methods operate at the token level, treating each output position independently and thus failing to concentrate the distillation signal on event-relevant spans or to capture the relational structure among event components. We propose EventKD, a two-level distillation framework tailored for event extraction. At the token level, a distributional divergence loss incorporates event-aware span importance weighting to prioritize tokens corresponding to triggers, arguments, and type labels. At the span level, a Relational-Aware (RA) loss preserves pairwise geometric relationships among event-aware spans across Transformer layers. Experiments on four benchmarks (ACE05, MAVEN, GENEVA, and RAMS) demonstrate that EventKD consistently outperforms existing distillation methods on both trigger identification and argument extraction.</abstract>
<identifier type="citekey">dat-etal-2026-eventkd</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.154/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>3240</start>
<end>3253</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T EventKD: Event-Aware Knowledge Distillation for Generative Event Extraction
%A Dat, Hoang Thanh
%A Dao, Quoc Phong
%A Hai, Nam Le
%A Dernoncourt, Franck
%A Van, Linh Ngo
%A Le, Trung
%A Nguyen, Thien Huu
%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 dat-etal-2026-eventkd
%X Event extraction aims to identify triggers in text, classify them into predefined event types, and recognize associated arguments such as participants and locations. Large language model (LLM)-based generative approaches provide a unified paradigm for joint trigger identification, classification, and argument extraction; however, their computational cost necessitates knowledge distillation (KD) into smaller student models. Existing KD methods operate at the token level, treating each output position independently and thus failing to concentrate the distillation signal on event-relevant spans or to capture the relational structure among event components. We propose EventKD, a two-level distillation framework tailored for event extraction. At the token level, a distributional divergence loss incorporates event-aware span importance weighting to prioritize tokens corresponding to triggers, arguments, and type labels. At the span level, a Relational-Aware (RA) loss preserves pairwise geometric relationships among event-aware spans across Transformer layers. Experiments on four benchmarks (ACE05, MAVEN, GENEVA, and RAMS) demonstrate that EventKD consistently outperforms existing distillation methods on both trigger identification and argument extraction.
%U https://aclanthology.org/2026.findings-emnlp.154/
%P 3240-3253
Markdown (Informal)
[EventKD: Event-Aware Knowledge Distillation for Generative Event Extraction](https://aclanthology.org/2026.findings-emnlp.154/) (Dat et al., Findings 2026)
ACL