Evaluating Neuron Interpretation Methods of NLP Models

Neuron Interpretation has gained traction in the field of interpretability, and have provided fine-grained insights into what a model learns and how language knowledge is distributed amongst its different components. However, the lack of evaluation benchmark and metrics have led to siloed progress w...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Fan, Yimin, Dalvi, Fahim, Durrani, Nadir, Sajjad, Hassan
Format: Artikel
Sprache:eng
Schlagworte:
Online-Zugang:Volltext bestellen
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
container_end_page
container_issue
container_start_page
container_title
container_volume
creator Fan, Yimin
Dalvi, Fahim
Durrani, Nadir
Sajjad, Hassan
description Neuron Interpretation has gained traction in the field of interpretability, and have provided fine-grained insights into what a model learns and how language knowledge is distributed amongst its different components. However, the lack of evaluation benchmark and metrics have led to siloed progress within these various methods, with very little work comparing them and highlighting their strengths and weaknesses. The reason for this discrepancy is the difficulty of creating ground truth datasets, for example, many neurons within a given model may learn the same phenomena, and hence there may not be one correct answer. Moreover, a learned phenomenon may spread across several neurons that work together -- surfacing these to create a gold standard challenging. In this work, we propose an evaluation framework that measures the compatibility of a neuron analysis method with other methods. We hypothesize that the more compatible a method is with the majority of the methods, the more confident one can be about its performance. We systematically evaluate our proposed framework and present a comparative analysis of a large set of neuron interpretation methods. We make the evaluation framework available to the community. It enables the evaluation of any new method using 20 concepts and across three pre-trained models.The code is released at https://github.com/fdalvi/neuron-comparative-analysis
doi_str_mv 10.48550/arxiv.2301.12608
format Article
fullrecord <record><control><sourceid>arxiv_GOX</sourceid><recordid>TN_cdi_arxiv_primary_2301_12608</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2301_12608</sourcerecordid><originalsourceid>FETCH-LOGICAL-a678-c4d1bdce8a83c28671e1ae9dc940b0fba4d186e8398c372685b8ea3a377262953</originalsourceid><addsrcrecordid>eNotj71ug0AQhK9xEWE_QKrcC4DvB46ltCwntgROCnq03C0JEgHrwJb99iEk1eibkUYzjD1LEcWQJGKL_t7eIqWFjKQyAp6YOdywu-LU9p_8TFc_9PzUT-QvnqbZnbGg6WtwIx8afs4_eDE46sY1WzXYjbT514CVr4dyfwzz97fTfpeHaFIIbexk7SwBgrYKTCpJImXOZrGoRVPjnIMh0BlYnSoDSQ2EGnU6g8oSHbCXv9pleHXx7Tf6R_V7oFoO6B__5z_X</addsrcrecordid><sourcetype>Open Access Repository</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype></control><display><type>article</type><title>Evaluating Neuron Interpretation Methods of NLP Models</title><source>arXiv.org</source><creator>Fan, Yimin ; Dalvi, Fahim ; Durrani, Nadir ; Sajjad, Hassan</creator><creatorcontrib>Fan, Yimin ; Dalvi, Fahim ; Durrani, Nadir ; Sajjad, Hassan</creatorcontrib><description>Neuron Interpretation has gained traction in the field of interpretability, and have provided fine-grained insights into what a model learns and how language knowledge is distributed amongst its different components. However, the lack of evaluation benchmark and metrics have led to siloed progress within these various methods, with very little work comparing them and highlighting their strengths and weaknesses. The reason for this discrepancy is the difficulty of creating ground truth datasets, for example, many neurons within a given model may learn the same phenomena, and hence there may not be one correct answer. Moreover, a learned phenomenon may spread across several neurons that work together -- surfacing these to create a gold standard challenging. In this work, we propose an evaluation framework that measures the compatibility of a neuron analysis method with other methods. We hypothesize that the more compatible a method is with the majority of the methods, the more confident one can be about its performance. We systematically evaluate our proposed framework and present a comparative analysis of a large set of neuron interpretation methods. We make the evaluation framework available to the community. It enables the evaluation of any new method using 20 concepts and across three pre-trained models.The code is released at https://github.com/fdalvi/neuron-comparative-analysis</description><identifier>DOI: 10.48550/arxiv.2301.12608</identifier><language>eng</language><subject>Computer Science - Computation and Language</subject><creationdate>2023-01</creationdate><rights>http://creativecommons.org/licenses/by/4.0</rights><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><link.rule.ids>228,230,780,885</link.rule.ids><linktorsrc>$$Uhttps://arxiv.org/abs/2301.12608$$EView_record_in_Cornell_University$$FView_record_in_$$GCornell_University$$Hfree_for_read</linktorsrc><backlink>$$Uhttps://doi.org/10.48550/arXiv.2301.12608$$DView paper in arXiv$$Hfree_for_read</backlink></links><search><creatorcontrib>Fan, Yimin</creatorcontrib><creatorcontrib>Dalvi, Fahim</creatorcontrib><creatorcontrib>Durrani, Nadir</creatorcontrib><creatorcontrib>Sajjad, Hassan</creatorcontrib><title>Evaluating Neuron Interpretation Methods of NLP Models</title><description>Neuron Interpretation has gained traction in the field of interpretability, and have provided fine-grained insights into what a model learns and how language knowledge is distributed amongst its different components. However, the lack of evaluation benchmark and metrics have led to siloed progress within these various methods, with very little work comparing them and highlighting their strengths and weaknesses. The reason for this discrepancy is the difficulty of creating ground truth datasets, for example, many neurons within a given model may learn the same phenomena, and hence there may not be one correct answer. Moreover, a learned phenomenon may spread across several neurons that work together -- surfacing these to create a gold standard challenging. In this work, we propose an evaluation framework that measures the compatibility of a neuron analysis method with other methods. We hypothesize that the more compatible a method is with the majority of the methods, the more confident one can be about its performance. We systematically evaluate our proposed framework and present a comparative analysis of a large set of neuron interpretation methods. We make the evaluation framework available to the community. It enables the evaluation of any new method using 20 concepts and across three pre-trained models.The code is released at https://github.com/fdalvi/neuron-comparative-analysis</description><subject>Computer Science - Computation and Language</subject><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2023</creationdate><recordtype>article</recordtype><sourceid>GOX</sourceid><recordid>eNotj71ug0AQhK9xEWE_QKrcC4DvB46ltCwntgROCnq03C0JEgHrwJb99iEk1eibkUYzjD1LEcWQJGKL_t7eIqWFjKQyAp6YOdywu-LU9p_8TFc_9PzUT-QvnqbZnbGg6WtwIx8afs4_eDE46sY1WzXYjbT514CVr4dyfwzz97fTfpeHaFIIbexk7SwBgrYKTCpJImXOZrGoRVPjnIMh0BlYnSoDSQ2EGnU6g8oSHbCXv9pleHXx7Tf6R_V7oFoO6B__5z_X</recordid><startdate>20230129</startdate><enddate>20230129</enddate><creator>Fan, Yimin</creator><creator>Dalvi, Fahim</creator><creator>Durrani, Nadir</creator><creator>Sajjad, Hassan</creator><scope>AKY</scope><scope>GOX</scope></search><sort><creationdate>20230129</creationdate><title>Evaluating Neuron Interpretation Methods of NLP Models</title><author>Fan, Yimin ; Dalvi, Fahim ; Durrani, Nadir ; Sajjad, Hassan</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-a678-c4d1bdce8a83c28671e1ae9dc940b0fba4d186e8398c372685b8ea3a377262953</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2023</creationdate><topic>Computer Science - Computation and Language</topic><toplevel>online_resources</toplevel><creatorcontrib>Fan, Yimin</creatorcontrib><creatorcontrib>Dalvi, Fahim</creatorcontrib><creatorcontrib>Durrani, Nadir</creatorcontrib><creatorcontrib>Sajjad, Hassan</creatorcontrib><collection>arXiv Computer Science</collection><collection>arXiv.org</collection></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext_linktorsrc</fulltext></delivery><addata><au>Fan, Yimin</au><au>Dalvi, Fahim</au><au>Durrani, Nadir</au><au>Sajjad, Hassan</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>Evaluating Neuron Interpretation Methods of NLP Models</atitle><date>2023-01-29</date><risdate>2023</risdate><abstract>Neuron Interpretation has gained traction in the field of interpretability, and have provided fine-grained insights into what a model learns and how language knowledge is distributed amongst its different components. However, the lack of evaluation benchmark and metrics have led to siloed progress within these various methods, with very little work comparing them and highlighting their strengths and weaknesses. The reason for this discrepancy is the difficulty of creating ground truth datasets, for example, many neurons within a given model may learn the same phenomena, and hence there may not be one correct answer. Moreover, a learned phenomenon may spread across several neurons that work together -- surfacing these to create a gold standard challenging. In this work, we propose an evaluation framework that measures the compatibility of a neuron analysis method with other methods. We hypothesize that the more compatible a method is with the majority of the methods, the more confident one can be about its performance. We systematically evaluate our proposed framework and present a comparative analysis of a large set of neuron interpretation methods. We make the evaluation framework available to the community. It enables the evaluation of any new method using 20 concepts and across three pre-trained models.The code is released at https://github.com/fdalvi/neuron-comparative-analysis</abstract><doi>10.48550/arxiv.2301.12608</doi><oa>free_for_read</oa></addata></record>
fulltext fulltext_linktorsrc
identifier DOI: 10.48550/arxiv.2301.12608
ispartof
issn
language eng
recordid cdi_arxiv_primary_2301_12608
source arXiv.org
subjects Computer Science - Computation and Language
title Evaluating Neuron Interpretation Methods of NLP Models
url https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-29T18%3A19%3A58IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-arxiv_GOX&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=Evaluating%20Neuron%20Interpretation%20Methods%20of%20NLP%20Models&rft.au=Fan,%20Yimin&rft.date=2023-01-29&rft_id=info:doi/10.48550/arxiv.2301.12608&rft_dat=%3Carxiv_GOX%3E2301_12608%3C/arxiv_GOX%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_id=info:pmid/&rfr_iscdi=true