@inproceedings{koulogeorge-etal-2025-bridging,
title = "Bridging the Faithfulness Gap in Prototypical Models",
author = "Koulogeorge, Andrew and
Xie, Sean and
Hassanpour, Saeed and
Vosoughi, Soroush",
editor = "Drozd, Aleksandr and
Sedoc, Jo{\~a}o and
Tafreshi, Shabnam and
Akula, Arjun and
Shu, Raphael",
booktitle = "The Sixth Workshop on Insights from Negative Results in NLP",
month = may,
year = "2025",
address = "Albuquerque, New Mexico",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.insights-1.9/",
doi = "10.18653/v1/2025.insights-1.9",
pages = "86--99",
ISBN = "979-8-89176-240-4",
abstract = "Prototypical Network-based Language Models (PNLMs) have been introduced as a novel approach for enhancing interpretability in deep learning models for NLP. In this work, we show that, despite the transparency afforded by their case-based reasoning architecture, current PNLMs are, in fact, not faithful, i.e. their explanations do not accurately reflect the underlying model{'}s reasoning process. By adopting an axiomatic approach grounded in the seminal works' definition of faithfulness, we identify two specific points in the architecture of PNLMs where unfaithfulness may occur. To address this, we introduce Faithful Alignment (FA), a two-part framework that ensures the faithfulness of PNLMs' explanations. We then demonstrate that FA achieves this goal without compromising model performance across a variety of downstream tasks and ablation studies."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="koulogeorge-etal-2025-bridging">
<titleInfo>
<title>Bridging the Faithfulness Gap in Prototypical Models</title>
</titleInfo>
<name type="personal">
<namePart type="given">Andrew</namePart>
<namePart type="family">Koulogeorge</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sean</namePart>
<namePart type="family">Xie</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Saeed</namePart>
<namePart type="family">Hassanpour</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Soroush</namePart>
<namePart type="family">Vosoughi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-05</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>The Sixth Workshop on Insights from Negative Results in NLP</title>
</titleInfo>
<name type="personal">
<namePart type="given">Aleksandr</namePart>
<namePart type="family">Drozd</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">João</namePart>
<namePart type="family">Sedoc</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shabnam</namePart>
<namePart type="family">Tafreshi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Arjun</namePart>
<namePart type="family">Akula</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Raphael</namePart>
<namePart type="family">Shu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Albuquerque, New Mexico</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-240-4</identifier>
</relatedItem>
<abstract>Prototypical Network-based Language Models (PNLMs) have been introduced as a novel approach for enhancing interpretability in deep learning models for NLP. In this work, we show that, despite the transparency afforded by their case-based reasoning architecture, current PNLMs are, in fact, not faithful, i.e. their explanations do not accurately reflect the underlying model’s reasoning process. By adopting an axiomatic approach grounded in the seminal works’ definition of faithfulness, we identify two specific points in the architecture of PNLMs where unfaithfulness may occur. To address this, we introduce Faithful Alignment (FA), a two-part framework that ensures the faithfulness of PNLMs’ explanations. We then demonstrate that FA achieves this goal without compromising model performance across a variety of downstream tasks and ablation studies.</abstract>
<identifier type="citekey">koulogeorge-etal-2025-bridging</identifier>
<identifier type="doi">10.18653/v1/2025.insights-1.9</identifier>
<location>
<url>https://aclanthology.org/2025.insights-1.9/</url>
</location>
<part>
<date>2025-05</date>
<extent unit="page">
<start>86</start>
<end>99</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Bridging the Faithfulness Gap in Prototypical Models
%A Koulogeorge, Andrew
%A Xie, Sean
%A Hassanpour, Saeed
%A Vosoughi, Soroush
%Y Drozd, Aleksandr
%Y Sedoc, João
%Y Tafreshi, Shabnam
%Y Akula, Arjun
%Y Shu, Raphael
%S The Sixth Workshop on Insights from Negative Results in NLP
%D 2025
%8 May
%I Association for Computational Linguistics
%C Albuquerque, New Mexico
%@ 979-8-89176-240-4
%F koulogeorge-etal-2025-bridging
%X Prototypical Network-based Language Models (PNLMs) have been introduced as a novel approach for enhancing interpretability in deep learning models for NLP. In this work, we show that, despite the transparency afforded by their case-based reasoning architecture, current PNLMs are, in fact, not faithful, i.e. their explanations do not accurately reflect the underlying model’s reasoning process. By adopting an axiomatic approach grounded in the seminal works’ definition of faithfulness, we identify two specific points in the architecture of PNLMs where unfaithfulness may occur. To address this, we introduce Faithful Alignment (FA), a two-part framework that ensures the faithfulness of PNLMs’ explanations. We then demonstrate that FA achieves this goal without compromising model performance across a variety of downstream tasks and ablation studies.
%R 10.18653/v1/2025.insights-1.9
%U https://aclanthology.org/2025.insights-1.9/
%U https://doi.org/10.18653/v1/2025.insights-1.9
%P 86-99
Markdown (Informal)
[Bridging the Faithfulness Gap in Prototypical Models](https://aclanthology.org/2025.insights-1.9/) (Koulogeorge et al., insights 2025)
ACL
- Andrew Koulogeorge, Sean Xie, Saeed Hassanpour, and Soroush Vosoughi. 2025. Bridging the Faithfulness Gap in Prototypical Models. In The Sixth Workshop on Insights from Negative Results in NLP, pages 86–99, Albuquerque, New Mexico. Association for Computational Linguistics.