@inproceedings{grecu-etal-2026-narsil,
title = "{N}ar{S}i{L} at {S}em{E}val-2026 Task 4: A Multi-Expert, Multi-Pathway System for Narrative Story Similarity",
author = "Grecu, Bogdan Octavian and
Chiru, Costin and
Cocarascu, Oana",
editor = "Kochmar, Ekaterina and
Ghosh, Debanjan and
North, Kai and
Komachi, Mamoru",
booktitle = "Proceedings of the 20th {I}nternational {W}orkshop on {S}emantic {E}valuation (2026)",
month = jul,
year = "2026",
address = "San Diego, California, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.semeval-1.381/",
pages = "3035--3044",
ISBN = "979-8-89176-414-9",
abstract = "We present NarSiL (Narrative Similarity Learners), our system for SemEval-2026 Task 4 Track A on Narrative Story Similarity. NarSiL employs a two-stage architecture: a Mixture-of-Experts (MoE) initial classifier that also leverages supermajority voting across three large language models (Gemma-3-12B, GPT-3.5-turbo-instruct, and Gemini-2.5-Flash) over multiple runs, followed by a structured three-pathway fallback for ambiguous cases. The three pathways correspond directly to the task{'}s three core similarity components, abstract theme, narrative outcome, and course of action. Each path yields a similarity score corresponding to its respective component, and the scores are then combined through a weighted aggregation step. NarSiL achieves 64.25{\%} accuracy on the official test set. An improved score of 70.25{\%} is obtained by considering only the supermajority voting of GPT, followed by the previously described fallback."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="grecu-etal-2026-narsil">
<titleInfo>
<title>NarSiL at SemEval-2026 Task 4: A Multi-Expert, Multi-Pathway System for Narrative Story Similarity</title>
</titleInfo>
<name type="personal">
<namePart type="given">Bogdan</namePart>
<namePart type="given">Octavian</namePart>
<namePart type="family">Grecu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Costin</namePart>
<namePart type="family">Chiru</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Oana</namePart>
<namePart type="family">Cocarascu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 20th International Workshop on Semantic Evaluation (2026)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ekaterina</namePart>
<namePart type="family">Kochmar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Debanjan</namePart>
<namePart type="family">Ghosh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kai</namePart>
<namePart type="family">North</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mamoru</namePart>
<namePart type="family">Komachi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, USA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-414-9</identifier>
</relatedItem>
<abstract>We present NarSiL (Narrative Similarity Learners), our system for SemEval-2026 Task 4 Track A on Narrative Story Similarity. NarSiL employs a two-stage architecture: a Mixture-of-Experts (MoE) initial classifier that also leverages supermajority voting across three large language models (Gemma-3-12B, GPT-3.5-turbo-instruct, and Gemini-2.5-Flash) over multiple runs, followed by a structured three-pathway fallback for ambiguous cases. The three pathways correspond directly to the task’s three core similarity components, abstract theme, narrative outcome, and course of action. Each path yields a similarity score corresponding to its respective component, and the scores are then combined through a weighted aggregation step. NarSiL achieves 64.25% accuracy on the official test set. An improved score of 70.25% is obtained by considering only the supermajority voting of GPT, followed by the previously described fallback.</abstract>
<identifier type="citekey">grecu-etal-2026-narsil</identifier>
<location>
<url>https://aclanthology.org/2026.semeval-1.381/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>3035</start>
<end>3044</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T NarSiL at SemEval-2026 Task 4: A Multi-Expert, Multi-Pathway System for Narrative Story Similarity
%A Grecu, Bogdan Octavian
%A Chiru, Costin
%A Cocarascu, Oana
%Y Kochmar, Ekaterina
%Y Ghosh, Debanjan
%Y North, Kai
%Y Komachi, Mamoru
%S Proceedings of the 20th International Workshop on Semantic Evaluation (2026)
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, USA
%@ 979-8-89176-414-9
%F grecu-etal-2026-narsil
%X We present NarSiL (Narrative Similarity Learners), our system for SemEval-2026 Task 4 Track A on Narrative Story Similarity. NarSiL employs a two-stage architecture: a Mixture-of-Experts (MoE) initial classifier that also leverages supermajority voting across three large language models (Gemma-3-12B, GPT-3.5-turbo-instruct, and Gemini-2.5-Flash) over multiple runs, followed by a structured three-pathway fallback for ambiguous cases. The three pathways correspond directly to the task’s three core similarity components, abstract theme, narrative outcome, and course of action. Each path yields a similarity score corresponding to its respective component, and the scores are then combined through a weighted aggregation step. NarSiL achieves 64.25% accuracy on the official test set. An improved score of 70.25% is obtained by considering only the supermajority voting of GPT, followed by the previously described fallback.
%U https://aclanthology.org/2026.semeval-1.381/
%P 3035-3044
Markdown (Informal)
[NarSiL at SemEval-2026 Task 4: A Multi-Expert, Multi-Pathway System for Narrative Story Similarity](https://aclanthology.org/2026.semeval-1.381/) (Grecu et al., SemEval 2026)
ACL