{"id":"https://openalex.org/W4365597196","doi":"https://doi.org/10.48550/arxiv.2304.05969","title":"Localizing Model Behavior with Path Patching","display_name":"Localizing Model Behavior with Path Patching","publication_year":2023,"publication_date":"2023-04-12","ids":{"openalex":"https://openalex.org/W4365597196","doi":"https://doi.org/10.48550/arxiv.2304.05969"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2304.05969","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2304.05969","pdf_url":"https://arxiv.org/pdf/2304.05969","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2304.05969","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5026166476","display_name":"Nicholas Goldowsky-Dill","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Goldowsky-Dill, Nicholas","raw_affiliation_strings":["Redwood Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Redwood Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":null,"display_name":"MacLeod, Chris","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"MacLeod, Chris","raw_affiliation_strings":["Redwood Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Redwood Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102509277","display_name":"Lucas Sato","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sato, Lucas","raw_affiliation_strings":["Redwood Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Redwood Research","institution_ids":[]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5082261951","display_name":"Aryaman Arora","orcid":"https://orcid.org/0000-0002-4977-8206","observed_orcids":["https://orcid.org/0000-0002-4977-8206"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Arora, Aryaman","raw_affiliation_strings":["Redwood Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Redwood Research","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":5,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.9970999956130981,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.9970999956130981,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.000699999975040555,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.00039999998989515007,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/induction-heads","display_name":"induction heads","score":0.7850000262260437},{"id":"https://openalex.org/keywords/gpt-2","display_name":"GPT-2","score":0.9779999852180481},{"id":"https://openalex.org/keywords/hypothesis-testing","display_name":"hypothesis testing","score":0.4350000023841858},{"id":"https://openalex.org/keywords/failure-modes","display_name":"failure modes","score":0.5889999866485596},{"id":"https://openalex.org/keywords/interpretability","display_name":"interpretability","score":0.31200000643730164}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7249678373336792},{"id":"https://openalex.org/C2777735758","wikidata":"https://www.wikidata.org/wiki/Q817765","display_name":"Path (computing)","level":2,"score":0.6808172464370728},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.6637681722640991},{"id":"https://openalex.org/C2777212361","wikidata":"https://www.wikidata.org/wiki/Q5127848","display_name":"Class (philosophy)","level":2,"score":0.5850787162780762},{"id":"https://openalex.org/C2776608160","wikidata":"https://www.wikidata.org/wiki/Q4785462","display_name":"Natural (archaeology)","level":2,"score":0.4356520175933838},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.41244417428970337},{"id":"https://openalex.org/C31258907","wikidata":"https://www.wikidata.org/wiki/Q1301371","display_name":"Computer network","level":1,"score":0.1291566789150238},{"id":"https://openalex.org/C95457728","wikidata":"https://www.wikidata.org/wiki/Q309","display_name":"History","level":0,"score":0.0},{"id":"https://openalex.org/C166957645","wikidata":"https://www.wikidata.org/wiki/Q23498","display_name":"Archaeology","level":1,"score":0.0},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2304.05969","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2304.05969","pdf_url":"https://arxiv.org/pdf/2304.05969","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2304.05969","doi":"https://doi.org/10.48550/arxiv.2304.05969","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2304.05969","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2304.05969","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2304.05969","pdf_url":"https://arxiv.org/pdf/2304.05969","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4365597196.pdf","grobid_xml":"https://content.openalex.org/works/W4365597196.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4387210443","https://openalex.org/W4404783717","https://openalex.org/W7154577389","https://openalex.org/W7108913623","https://openalex.org/W2951257191","https://openalex.org/W6884719919","https://openalex.org/W7079293072","https://openalex.org/W2290102016","https://openalex.org/W6925680399","https://openalex.org/W7223113814"],"abstract_inverted_index":{"Localizing":[0],"behaviors":[1,70],"of":[2,8,15,66,76,82,88],"neural":[3],"networks":[4],"to":[5,48,73],"a":[6,13,20,56,63,74,86,93],"subset":[7,14],"the":[9,45],"network's":[10],"components":[11,18],"or":[12],"interactions":[16],"between":[17],"is":[19,34,41],"natural":[21,64],"first":[22],"step":[23],"towards":[24],"analyzing":[25],"network":[26],"mechanisms":[27],"and":[28,37,39,60,90],"possible":[29],"failure":[30],"modes.":[31],"Existing":[32],"work":[33],"often":[35],"qualitative":[36],"ad-hoc,":[38],"there":[40],"no":[42],"consensus":[43],"on":[44],"appropriate":[46],"way":[47],"evaluate":[49],"localization":[50],"claims.":[51],"We":[52,78],"introduce":[53],"path":[54],"patching,":[55],"technique":[57],"for":[58,95],"expressing":[59,68],"quantitatively":[61],"testing":[62],"class":[65],"hypotheses":[67],"that":[69],"are":[71],"localized":[72],"set":[75],"paths.":[77],"refine":[79],"an":[80],"explanation":[81],"induction":[83],"heads,":[84],"characterize":[85],"behavior":[87],"GPT-2,":[89],"open":[91],"source":[92],"framework":[94],"efficiently":[96],"running":[97],"similar":[98],"experiments.":[99]},"counts_by_year":[{"year":2026,"cited_by_count":2},{"year":2025,"cited_by_count":2},{"year":2023,"cited_by_count":1}],"updated_date":"2026-10-16T09:59:22.235173","created_date":"2023-04-15T00:00:00"}