{"id":"https://openalex.org/W4412945693","doi":"https://doi.org/10.18653/v1/2025.acl-long.288","title":"Enhancing Automated Interpretability with Output-Centric Feature Descriptions","display_name":"Enhancing Automated Interpretability with Output-Centric Feature Descriptions","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412945693","doi":"https://doi.org/10.18653/v1/2025.acl-long.288"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.acl-long.288","doi":"https://doi.org/10.18653/v1/2025.acl-long.288","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.288","pdf_url":null,"source":{"id":"https://openalex.org/S4393915740","display_name":"Proceedings of the conference - Association for Computational Linguistics. Meeting","issn_l":"0736-587X","issn":["0736-587X"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":["norway-1"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://doi.org/10.18653/v1/2025.acl-long.288","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5115904413","display_name":"Yoav Gur-Arieh","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yoav Gur-Arieh","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5115904414","display_name":"Roy Mayan","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Roy Mayan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5115904415","display_name":"Chen Agassy","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen Agassy","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5002577142","display_name":"Atticus Geiger","orcid":"https://orcid.org/0000-0002-9170-506X","observed_orcids":["https://orcid.org/0000-0002-9170-506X"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Atticus Geiger","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5065717258","display_name":"Mor Geva","orcid":"https://orcid.org/0000-0001-9529-6315","observed_orcids":["https://orcid.org/0000-0001-9529-6315"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mor Geva","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.9496,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.71996595,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":90,"max":94},"biblio":{"volume":null,"issue":null,"first_page":"5757","last_page":"5778"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.9994999766349792,"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":0.9994999766349792,"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/T10743","display_name":"Software Testing and Debugging Techniques","score":9.999999747378752e-05,"subfield":{"id":"https://openalex.org/subfields/1712","display_name":"Software"},"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/T10260","display_name":"Software Engineering Research","score":0.0,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/explainable-ai","display_name":"explainable AI","score":0.6919999718666077},{"id":"https://openalex.org/keywords/feature-description","display_name":"feature description","score":0.4779999852180481},{"id":"https://openalex.org/keywords/model-interpretability","display_name":"model interpretability","score":0.5740000009536743}],"concepts":[{"id":"https://openalex.org/C2781067378","wikidata":"https://www.wikidata.org/wiki/Q17027399","display_name":"Interpretability","level":2,"score":0.9355601668357849},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7294234037399292},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.6462903618812561},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4835236072540283},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.362729549407959},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.33097317814826965},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.0},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.0}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/v1/2025.acl-long.288","doi":"https://doi.org/10.18653/v1/2025.acl-long.288","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.288","pdf_url":null,"source":{"id":"https://openalex.org/S4393915740","display_name":"Proceedings of the conference - Association for Computational Linguistics. Meeting","issn_l":"0736-587X","issn":["0736-587X"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":["norway-1"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2501.08319","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2501.08319","pdf_url":"https://arxiv.org/pdf/2501.08319","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2501.08319","doi":"https://doi.org/10.48550/arxiv.2501.08319","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2501.08319","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.acl-long.288","doi":"https://doi.org/10.18653/v1/2025.acl-long.288","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.288","pdf_url":null,"source":{"id":"https://openalex.org/S4393915740","display_name":"Proceedings of the conference - Association for Computational Linguistics. Meeting","issn_l":"0736-587X","issn":["0736-587X"],"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":["norway-1"],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4412887961","https://openalex.org/W4417357092","https://openalex.org/W4415314968","https://openalex.org/W7160261982","https://openalex.org/W4403995319","https://openalex.org/W4416232592","https://openalex.org/W4416037591","https://openalex.org/W7153217318","https://openalex.org/W7126375801","https://openalex.org/W7161099702"],"abstract_inverted_index":{"Automated":[0],"interpretability":[1],"pipelines":[2,92],"generate":[3],"natural":[4],"language":[5,15],"descriptions":[6,29,94,147,182],"for":[7,115],"the":[8,22,36,47,58,99,103,123,131,137,143,150,164,168],"concepts":[9],"represented":[10],"by":[11,70,79],"features":[12,191],"in":[13,25,46,64],"large":[14],"models":[16],"(LLMs),":[17],"such":[18],"as":[19],"plants":[20],"or":[21,43,130],"first":[23],"word":[24],"a":[26,41,44,62,74,154],"sentence.":[27],"These":[28,120],"are":[30],"derived":[31],"using":[32],"inputs":[33,54,72,188],"that":[34,90,95,180,189],"activate":[35,77,190],"feature,":[37],"which":[38],"may":[39],"be":[40,184,195],"dimension":[42],"direction":[45],"model's":[48],"representation":[49],"space.":[50],"However,":[51],"identifying":[52],"activating":[53],"is":[55,67],"costly,":[56],"and":[57,78,174],"mechanistic":[59],"role":[60],"of":[61,102,153],"feature":[63,75,81,104,118,128,155],"model":[65,157],"behavior":[66],"determined":[68],"both":[69,172],"how":[71,80],"cause":[73],"to":[76,97,142,167,186,194],"activation":[82],"affects":[83],"outputs.":[84,106],"Using":[85],"steering":[86],"evaluations,":[87],"we":[88,110,178],"reveal":[89],"current":[91],"provide":[93],"fail":[96],"capture":[98,149],"causal":[100,151],"effect":[101,152],"on":[105,156,171],"To":[107],"fix":[108],"this,":[109],"propose":[111],"efficient,":[112],"output-centric":[113,146,181],"methods":[114,121],"automatically":[116],"generating":[117],"descriptions.":[119],"use":[122],"tokens":[124,134],"weighted":[125],"higher":[126],"after":[127,135],"stimulation":[129],"highest":[132],"weight":[133],"applying":[136],"vocabulary":[138],"\"unembedding\"":[139],"head":[140],"directly":[141],"feature.":[144],"Our":[145],"better":[148],"outputs":[158],"than":[159],"input-centric":[160],"descriptions,":[161],"but":[162],"combining":[163],"two":[165],"leads":[166],"best":[169],"performance":[170],"input":[173],"output":[175],"evaluations.":[176],"Lastly,":[177],"show":[179],"can":[183],"used":[185],"find":[187],"previously":[192],"thought":[193],"\"dead\".":[196]},"counts_by_year":[{"year":2025,"cited_by_count":1}],"updated_date":"2026-10-29T09:56:32.521200","created_date":"2025-10-10T00:00:00"}