Unveiling Transformer Perception by Exploring Input Manifolds

This paper introduces a general method for the exploration of equivalence classes in the input space of Transformer models. The proposed approach is based on sound mathematical theory which describes the internal layers of a Transformer architecture as sequential deformations of the input manifold....

Full description

Saved in:
Bibliographic Details
Published in:arXiv.org 2024-10
Main Authors: Benfenati, Alessandro, Ferrara, Alfio, Alessio, Marta, Riva, Davide, Rocchetti, Elisabetta
Format: Article
Language:English
Subjects:
Computer vision
Equivalence
Natural language processing
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
cited_by
cites
container_end_page
container_issue
container_start_page
container_title arXiv.org
container_volume
creator Benfenati, Alessandro
Ferrara, Alfio
Alessio, Marta
Riva, Davide
Rocchetti, Elisabetta
description This paper introduces a general method for the exploration of equivalence classes in the input space of Transformer models. The proposed approach is based on sound mathematical theory which describes the internal layers of a Transformer architecture as sequential deformations of the input manifold. Using eigendecomposition of the pullback of the distance metric defined on the output space through the Jacobian of the model, we are able to reconstruct equivalence classes in the input space and navigate across them. We illustrate how this method can be used as a powerful tool for investigating how a Transformer sees the input space, facilitating local and task-agnostic explainability in Computer Vision and Natural Language Processing tasks.
format article
fullrecord <record><control><sourceid>proquest</sourceid><recordid>TN_cdi_proquest_journals_3115226943</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>3115226943</sourcerecordid><originalsourceid>FETCH-proquest_journals_31152269433</originalsourceid><addsrcrecordid>eNpjYuA0MjY21LUwMTLiYOAtLs4yMDAwMjM3MjU15mSwDc0rS83MycxLVwgpSswrTssvyk0tUghILUpOLSjJzM9TSKpUcK0oyMkvAqnxzCsoLVHwTczLTMvPSSnmYWBNS8wpTuWF0twMym6uIc4eugVF-YWlqcUl8Vn5pUV5QKl4Y0NDUyMjM0sTY2PiVAEAvSc4Zw</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>3115226943</pqid></control><display><type>article</type><title>Unveiling Transformer Perception by Exploring Input Manifolds</title><source>Publicly Available Content Database</source><creator>Benfenati, Alessandro ; Ferrara, Alfio ; Alessio, Marta ; Riva, Davide ; Rocchetti, Elisabetta</creator><creatorcontrib>Benfenati, Alessandro ; Ferrara, Alfio ; Alessio, Marta ; Riva, Davide ; Rocchetti, Elisabetta</creatorcontrib><description>This paper introduces a general method for the exploration of equivalence classes in the input space of Transformer models. The proposed approach is based on sound mathematical theory which describes the internal layers of a Transformer architecture as sequential deformations of the input manifold. Using eigendecomposition of the pullback of the distance metric defined on the output space through the Jacobian of the model, we are able to reconstruct equivalence classes in the input space and navigate across them. We illustrate how this method can be used as a powerful tool for investigating how a Transformer sees the input space, facilitating local and task-agnostic explainability in Computer Vision and Natural Language Processing tasks.</description><identifier>EISSN: 2331-8422</identifier><language>eng</language><publisher>Ithaca: Cornell University Library, arXiv.org</publisher><subject>Computer vision ; Equivalence ; Natural language processing</subject><ispartof>arXiv.org, 2024-10</ispartof><rights>2024. This work is published under http://creativecommons.org/licenses/by-sa/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktohtml>$$Uhttps://www.proquest.com/docview/3115226943?pq-origsite=primo$$EHTML$$P50$$Gproquest$$Hfree_for_read</linktohtml><link.rule.ids>776,780,25731,36989,44566</link.rule.ids></links><search><creatorcontrib>Benfenati, Alessandro</creatorcontrib><creatorcontrib>Ferrara, Alfio</creatorcontrib><creatorcontrib>Alessio, Marta</creatorcontrib><creatorcontrib>Riva, Davide</creatorcontrib><creatorcontrib>Rocchetti, Elisabetta</creatorcontrib><title>Unveiling Transformer Perception by Exploring Input Manifolds</title><title>arXiv.org</title><description>This paper introduces a general method for the exploration of equivalence classes in the input space of Transformer models. The proposed approach is based on sound mathematical theory which describes the internal layers of a Transformer architecture as sequential deformations of the input manifold. Using eigendecomposition of the pullback of the distance metric defined on the output space through the Jacobian of the model, we are able to reconstruct equivalence classes in the input space and navigate across them. We illustrate how this method can be used as a powerful tool for investigating how a Transformer sees the input space, facilitating local and task-agnostic explainability in Computer Vision and Natural Language Processing tasks.</description><subject>Computer vision</subject><subject>Equivalence</subject><subject>Natural language processing</subject><issn>2331-8422</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2024</creationdate><recordtype>article</recordtype><sourceid>PIMPY</sourceid><recordid>eNpjYuA0MjY21LUwMTLiYOAtLs4yMDAwMjM3MjU15mSwDc0rS83MycxLVwgpSswrTssvyk0tUghILUpOLSjJzM9TSKpUcK0oyMkvAqnxzCsoLVHwTczLTMvPSSnmYWBNS8wpTuWF0twMym6uIc4eugVF-YWlqcUl8Vn5pUV5QKl4Y0NDUyMjM0sTY2PiVAEAvSc4Zw</recordid><startdate>20241008</startdate><enddate>20241008</enddate><creator>Benfenati, Alessandro</creator><creator>Ferrara, Alfio</creator><creator>Alessio, Marta</creator><creator>Riva, Davide</creator><creator>Rocchetti, Elisabetta</creator><general>Cornell University Library, arXiv.org</general><scope>8FE</scope><scope>8FG</scope><scope>ABJCF</scope><scope>ABUWG</scope><scope>AFKRA</scope><scope>AZQEC</scope><scope>BENPR</scope><scope>BGLVJ</scope><scope>CCPQU</scope><scope>DWQXO</scope><scope>HCIFZ</scope><scope>L6V</scope><scope>M7S</scope><scope>PIMPY</scope><scope>PQEST</scope><scope>PQQKQ</scope><scope>PQUKI</scope><scope>PRINS</scope><scope>PTHSS</scope></search><sort><creationdate>20241008</creationdate><title>Unveiling Transformer Perception by Exploring Input Manifolds</title><author>Benfenati, Alessandro ; Ferrara, Alfio ; Alessio, Marta ; Riva, Davide ; Rocchetti, Elisabetta</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-proquest_journals_31152269433</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2024</creationdate><topic>Computer vision</topic><topic>Equivalence</topic><topic>Natural language processing</topic><toplevel>online_resources</toplevel><creatorcontrib>Benfenati, Alessandro</creatorcontrib><creatorcontrib>Ferrara, Alfio</creatorcontrib><creatorcontrib>Alessio, Marta</creatorcontrib><creatorcontrib>Riva, Davide</creatorcontrib><creatorcontrib>Rocchetti, Elisabetta</creatorcontrib><collection>ProQuest SciTech Collection</collection><collection>ProQuest Technology Collection</collection><collection>Materials Science &amp; Engineering Collection</collection><collection>ProQuest Central (Alumni)</collection><collection>ProQuest Central</collection><collection>ProQuest Central Essentials</collection><collection>ProQuest Central</collection><collection>Technology Collection</collection><collection>ProQuest One Community College</collection><collection>ProQuest Central</collection><collection>SciTech Premium Collection</collection><collection>ProQuest Engineering Collection</collection><collection>Engineering Database</collection><collection>Publicly Available Content Database</collection><collection>ProQuest One Academic Eastern Edition (DO NOT USE)</collection><collection>ProQuest One Academic</collection><collection>ProQuest One Academic UKI Edition</collection><collection>ProQuest Central China</collection><collection>Engineering Collection</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Benfenati, Alessandro</au><au>Ferrara, Alfio</au><au>Alessio, Marta</au><au>Riva, Davide</au><au>Rocchetti, Elisabetta</au><format>book</format><genre>document</genre><ristype>GEN</ristype><atitle>Unveiling Transformer Perception by Exploring Input Manifolds</atitle><jtitle>arXiv.org</jtitle><date>2024-10-08</date><risdate>2024</risdate><eissn>2331-8422</eissn><abstract>This paper introduces a general method for the exploration of equivalence classes in the input space of Transformer models. The proposed approach is based on sound mathematical theory which describes the internal layers of a Transformer architecture as sequential deformations of the input manifold. Using eigendecomposition of the pullback of the distance metric defined on the output space through the Jacobian of the model, we are able to reconstruct equivalence classes in the input space and navigate across them. We illustrate how this method can be used as a powerful tool for investigating how a Transformer sees the input space, facilitating local and task-agnostic explainability in Computer Vision and Natural Language Processing tasks.</abstract><cop>Ithaca</cop><pub>Cornell University Library, arXiv.org</pub><oa>free_for_read</oa></addata></record>
fulltext fulltext
identifier EISSN: 2331-8422
ispartof arXiv.org, 2024-10
issn 2331-8422
language eng
recordid cdi_proquest_journals_3115226943
source Publicly Available Content Database
subjects Computer vision
Equivalence
Natural language processing
title Unveiling Transformer Perception by Exploring Input Manifolds
url http://sfxeu10.hosted.exlibrisgroup.com/loughborough?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2025-01-30T06%3A59%3A52IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest&rft_val_fmt=info:ofi/fmt:kev:mtx:book&rft.genre=document&rft.atitle=Unveiling%20Transformer%20Perception%20by%20Exploring%20Input%20Manifolds&rft.jtitle=arXiv.org&rft.au=Benfenati,%20Alessandro&rft.date=2024-10-08&rft.eissn=2331-8422&rft_id=info:doi/&rft_dat=%3Cproquest%3E3115226943%3C/proquest%3E%3Cgrp_id%3Ecdi_FETCH-proquest_journals_31152269433%3C/grp_id%3E%3Coa%3E%3C/oa%3E%3Curl%3E%3C/url%3E&rft_id=info:oai/&rft_pqid=3115226943&rft_id=info:pmid/&rfr_iscdi=true