{"id":"https://openalex.org/W7166858631","doi":"https://doi.org/10.18653/v1/2026.acl-long.154","title":"Multi-component Causal Tracing in Large Language Models","display_name":"Multi-component Causal Tracing in Large Language Models","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7166858631","doi":"https://doi.org/10.18653/v1/2026.acl-long.154"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2026.acl-long.154","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.154","pdf_url":"https://aclanthology.org/2026.acl-long.154.pdf","source":null,"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 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.acl-long.154.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5043350420","display_name":"Zirui Yan","orcid":"https://orcid.org/0009-0003-1314-7773"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zirui Yan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139723981","display_name":"Dennis Wei","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dennis Wei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137753860","display_name":"Dmitriy A. Katz","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dmitriy A Katz","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5060534465","display_name":"Prasanna Sattigeri","orcid":"https://orcid.org/0000-0003-4435-0486"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Prasanna Sattigeri","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5054103414","display_name":"Ali Tajer","orcid":"https://orcid.org/0000-0002-3513-4135"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ali Tajer","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":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.84711139,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"3398","last_page":"3418"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11303","display_name":"Bayesian Modeling and Causal Inference","score":0.5242999792098999,"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/T11303","display_name":"Bayesian Modeling and Causal Inference","score":0.5242999792098999,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.0714000016450882,"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/T10845","display_name":"Advanced Causal Inference Techniques","score":0.056699998676776886,"subfield":{"id":"https://openalex.org/subfields/2613","display_name":"Statistics and Probability"},"field":{"id":"https://openalex.org/fields/26","display_name":"Mathematics"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/tracing","display_name":"Tracing","score":0.43560001254081726},{"id":"https://openalex.org/keywords/natural-language","display_name":"Natural language","score":0.329800009727478},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.32019999623298645},{"id":"https://openalex.org/keywords/causal-model","display_name":"Causal model","score":0.3167000114917755},{"id":"https://openalex.org/keywords/causality","display_name":"Causality (physics)","score":0.26579999923706055}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6114000082015991},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.44020000100135803},{"id":"https://openalex.org/C138673069","wikidata":"https://www.wikidata.org/wiki/Q322229","display_name":"Tracing","level":2,"score":0.43560001254081726},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3935000002384186},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.329800009727478},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.32019999623298645},{"id":"https://openalex.org/C11671645","wikidata":"https://www.wikidata.org/wiki/Q5054567","display_name":"Causal model","level":2,"score":0.3167000114917755},{"id":"https://openalex.org/C41895202","wikidata":"https://www.wikidata.org/wiki/Q8162","display_name":"Linguistics","level":1,"score":0.28529998660087585},{"id":"https://openalex.org/C64357122","wikidata":"https://www.wikidata.org/wiki/Q1149766","display_name":"Causality (physics)","level":2,"score":0.26579999923706055},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2443999946117401}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.acl-long.154","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.154","pdf_url":"https://aclanthology.org/2026.acl-long.154.pdf","source":null,"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 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.acl-long.154","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.acl-long.154","pdf_url":"https://aclanthology.org/2026.acl-long.154.pdf","source":null,"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 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[{"score":0.4417976438999176,"display_name":"Reduced inequalities","id":"https://metadata.un.org/sdg/10"}],"awards":[],"funders":[{"id":"https://openalex.org/F4320306076","display_name":"National Science Foundation","ror":"https://ror.org/021nxhr62"},{"id":"https://openalex.org/F4320309500","display_name":"Rensselaer Polytechnic Institute","ror":"https://ror.org/01rtyzb94"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7166858631.pdf","grobid_xml":"https://content.openalex.org/works/W7166858631.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Causal":[0],"tracing":[1,47],"systematically":[2,52],"intervenes":[3],"on":[4,33,157],"a":[5,42,68,85,109,120,154],"large":[6],"language":[7],"model's":[8,150],"(LLM's)":[9],"internal":[10],"representations":[11],"to":[12,24,67,84],"uncover":[13],"and":[14,61,75,108],"quantify":[15],"the":[16,30,54,92,96,115,142,149,158],"causal":[17],"pathways":[18],"linking":[19],"specific":[20,25],"inputs":[21],"or":[22,36],"computations":[23],"metrics":[26],"of":[27,56,88,95,148],"interest,":[28],"quantifying":[29],"LLM's":[31],"behavior.Building":[32],"previous":[34],"single-component":[35],"single-layer":[37],"studies,":[38],"this":[39],"paper":[40],"presents":[41],"unified":[43],"framework":[44,51],"for":[45,136],"causally":[46],"multiple":[48],"components":[49,57,151],"simultaneously.This":[50],"identifies":[53,146],"subsets":[55,147],"(e.g.,":[58,73],"attention":[59],"heads":[60],"multi-layer":[62],"perceptron":[63],"neurons)":[64],"most":[65],"critical":[66],"desired":[69,89],"target":[70,159],"performance":[71],"metric":[72,112],"accuracy":[74],"fairness).This":[76],"is":[77,102,166],"achieved":[78],"by":[79],"incorporating":[80],"flexible":[81],"interventions":[82,107],"applied":[83],"wide":[86],"range":[87],"metrics.To":[90],"address":[91],"combinatorial":[93,116],"complexity":[94],"multicomponent":[97],"problem,":[98],"an":[99],"efficient":[100],"algorithm":[101],"designed":[103,111],"that":[104,123,141,152],"leverages":[105],"soft":[106],"carefully":[110],"transformation,":[113],"converting":[114],"search":[117],"problem":[118],"into":[119],"continuous":[121],"one":[122],"can":[124],"be":[125],"solved":[126],"efficiently":[127,145],"under":[128],"proper":[129,133],"constraints,":[130],"thereby":[131],"generating":[132],"binary":[134],"decisions":[135],"selecting":[137],"components.Experimental":[138],"results":[139],"demonstrate":[140],"proposed":[143],"method":[144],"have":[153],"high":[155],"impact":[156],"metric,":[160],"outperforming":[161],"existing":[162],"baseline":[163],"approaches.Our":[164],"code":[165],"available":[167],"at":[168],"https://github.com/ZiruiYan/":[169],"multi-component-causal-tracing.":[170]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-07-02T00:00:00"}