{"id":"https://openalex.org/W4401661463","doi":"https://doi.org/10.1145/3677052.3698612","title":"Mechanistic interpretability of large language models with applications to the financial services industry","display_name":"Mechanistic interpretability of large language models with applications to the financial services industry","publication_year":2024,"publication_date":"2024-11-14","ids":{"openalex":"https://openalex.org/W4401661463","doi":"https://doi.org/10.1145/3677052.3698612"},"language":"en","primary_location":{"id":"doi:10.1145/3677052.3698612","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3677052.3698612","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3677052.3698612","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 5th ACM International Conference on AI in Finance","raw_type":"proceedings-article"},"type":"preprint","indexed_in":["arxiv","crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://dl.acm.org/doi/pdf/10.1145/3677052.3698612","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5106580061","display_name":"Ashkan Golgoon","orcid":"https://orcid.org/0000-0002-0251-4203"},"institutions":[{"id":"https://openalex.org/I4401726971","display_name":"Discover Financial Services (United States)","ror":"https://ror.org/05783n648","country_code":null,"type":"company","lineage":["https://openalex.org/I4401726971"]}],"countries":["US"],"is_corresponding":true,"raw_author_name":"Ashkan Golgoon","raw_affiliation_strings":["Discover Financial Services, USA"],"raw_orcid":"https://orcid.org/0000-0002-0251-4203","affiliations":[{"raw_affiliation_string":"Discover Financial Services, USA","institution_ids":["https://openalex.org/I4401726971"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5009622310","display_name":"Khashayar Filom","orcid":"https://orcid.org/0000-0002-6881-4460"},"institutions":[{"id":"https://openalex.org/I4401726971","display_name":"Discover Financial Services (United States)","ror":"https://ror.org/05783n648","country_code":null,"type":"company","lineage":["https://openalex.org/I4401726971"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Khashayar Filom","raw_affiliation_strings":["Discover Financial Services, USA"],"raw_orcid":"https://orcid.org/0000-0002-6881-4460","affiliations":[{"raw_affiliation_string":"Discover Financial Services, USA","institution_ids":["https://openalex.org/I4401726971"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5089840355","display_name":"Arjun Ravi Kannan","orcid":"https://orcid.org/0000-0003-4498-1800"},"institutions":[{"id":"https://openalex.org/I4401726971","display_name":"Discover Financial Services (United States)","ror":"https://ror.org/05783n648","country_code":null,"type":"company","lineage":["https://openalex.org/I4401726971"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Arjun Ravi Kannan","raw_affiliation_strings":["Discover Financial Services, USA"],"raw_orcid":"https://orcid.org/0000-0003-4498-1800","affiliations":[{"raw_affiliation_string":"Discover Financial Services, USA","institution_ids":["https://openalex.org/I4401726971"]}]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":3,"corresponding_author_ids":["https://openalex.org/A5106580061"],"corresponding_institution_ids":["https://openalex.org/I4401726971"],"apc_list":null,"apc_paid":null,"fwci":1.6557,"has_fulltext":true,"cited_by_count":5,"citation_normalized_percentile":{"value":0.86303112,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":96,"max":99},"biblio":{"volume":null,"issue":null,"first_page":"660","last_page":"668"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10456","display_name":"Multi-Agent Systems and Negotiation","score":0.2946999967098236,"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/T10456","display_name":"Multi-Agent Systems and Negotiation","score":0.2946999967098236,"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/interpretability","display_name":"Interpretability","score":0.9530254602432251},{"id":"https://openalex.org/keywords/financial-services","display_name":"Financial services","score":0.6791337728500366},{"id":"https://openalex.org/keywords/business","display_name":"Business","score":0.44170263409614563},{"id":"https://openalex.org/keywords/financial-modeling","display_name":"Financial modeling","score":0.4102582633495331},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.3794441819190979},{"id":"https://openalex.org/keywords/finance","display_name":"Finance","score":0.3677715063095093},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.2954137325286865}],"concepts":[{"id":"https://openalex.org/C2781067378","wikidata":"https://www.wikidata.org/wiki/Q17027399","display_name":"Interpretability","level":2,"score":0.9530254602432251},{"id":"https://openalex.org/C139043278","wikidata":"https://www.wikidata.org/wiki/Q837171","display_name":"Financial services","level":2,"score":0.6791337728500366},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.44170263409614563},{"id":"https://openalex.org/C23925645","wikidata":"https://www.wikidata.org/wiki/Q5449731","display_name":"Financial modeling","level":2,"score":0.4102582633495331},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.3794441819190979},{"id":"https://openalex.org/C10138342","wikidata":"https://www.wikidata.org/wiki/Q43015","display_name":"Finance","level":1,"score":0.3677715063095093},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.2954137325286865}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1145/3677052.3698612","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3677052.3698612","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3677052.3698612","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 5th ACM International Conference on AI in Finance","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2407.11215","is_oa":true,"landing_page_url":"http://arxiv.org/abs/2407.11215","pdf_url":"https://arxiv.org/pdf/2407.11215","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"is_oa":true,"is_in_doaj":false,"is_core":false,"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"}],"best_oa_location":{"id":"doi:10.1145/3677052.3698612","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3677052.3698612","pdf_url":"https://dl.acm.org/doi/pdf/10.1145/3677052.3698612","source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 5th ACM International Conference on AI in Finance","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":false},"content_urls":{"pdf":"https://content.openalex.org/works/W4401661463.pdf"},"referenced_works_count":9,"referenced_works":["https://openalex.org/W3010694149","https://openalex.org/W3110749958","https://openalex.org/W4234322080","https://openalex.org/W4308023630","https://openalex.org/W4367628394","https://openalex.org/W4388994228","https://openalex.org/W6739901393","https://openalex.org/W6839328737","https://openalex.org/W6843370592"],"related_works":["https://openalex.org/W4289731310","https://openalex.org/W2382239572","https://openalex.org/W4230959190","https://openalex.org/W4230534681","https://openalex.org/W2355382759","https://openalex.org/W2366257090","https://openalex.org/W4311705411","https://openalex.org/W3006132460","https://openalex.org/W2982265479","https://openalex.org/W4282592372"],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"exhibit":[3],"remarkable":[4],"capabilities":[5],"across":[6],"a":[7,168,203],"broad":[8],"spectrum":[9],"of":[10,31,55,78,89,103,128,140],"applications.":[11,98],"Nevertheless,":[12],"due":[13],"to":[14,39,81,124,148,172],"their":[15,25,40],"intrinsic":[16],"complexity,":[17],"these":[18],"models":[19,66,92],"present":[20],"substantial":[21],"challenges":[22,35],"in":[23,95,152,206],"interpreting":[24],"internal":[26],"decision-making":[27],"processes.":[28],"This":[29],"lack":[30],"transparency":[32],"poses":[33],"critical":[34],"when":[36,122],"it":[37],"comes":[38],"adaptation":[41],"by":[42],"financial":[43,96],"institutions,":[44],"where":[45],"concerns":[46],"and":[47,52,143,160,163,191,200],"accountability":[48],"regarding":[49],"bias,":[50],"fairness,":[51],"reliability":[53],"are":[54,74],"paramount":[56],"importance.":[57],"Mechanistic":[58],"interpretability":[59,80],"aims":[60],"at":[61],"reverse":[62],"engineering":[63],"complex":[64],"AI":[65],"such":[67],"as":[68,167,193,195],"transformers.":[69],"In":[70,114],"this":[71],"paper,":[72],"we":[73,116,136,157],"pioneering":[75],"the":[76,86,138,149,153,182,196,207],"use":[77,94,164],"mechanistic":[79],"shed":[82],"some":[83],"light":[84],"on":[85],"inner":[87],"workings":[88],"large":[90],"language":[91],"for":[93,110],"services":[97],"We":[99,179],"offer":[100],"several":[101],"examples":[102],"how":[104],"algorithmic":[105],"tasks":[106],"can":[107],"be":[108],"designed":[109],"compliance":[111],"monitoring":[112],"purposes.":[113],"particular,":[115],"investigate":[117],"GPT-2":[118],"Small\u2019s":[119],"attention":[120,146],"pattern":[121],"prompted":[123],"identify":[125],"potential":[126],"violation":[127],"Fair":[129],"Lending":[130],"laws.":[131],"Using":[132],"direct":[133],"logit":[134,150],"attribution,":[135],"study":[137],"contributions":[139],"each":[141],"layer":[142,188],"its":[144],"corresponding":[145],"heads":[147,184,198],"difference":[151],"residual":[154],"stream.":[155],"Finally,":[156],"design":[158],"clean":[159],"corrupted":[161],"prompts":[162],"activation":[165],"patching":[166],"causal":[169],"intervention":[170],"method":[171],"localize":[173],"our":[174],"task":[175,208],"completion":[176],"components":[177],"further.":[178],"observe":[180],"that":[181],"(positive)":[183],"10.2":[185],"(head":[186],"2,":[187],"10),":[189],"10.7,":[190],"11.3,":[192],"well":[194],"(negative)":[197],"9.6":[199],"10.6":[201],"play":[202],"significant":[204],"role":[205],"completion.":[209]},"counts_by_year":[{"year":2026,"cited_by_count":2},{"year":2025,"cited_by_count":3}],"updated_date":"2026-03-11T06:11:40.159057","created_date":"2024-08-17T00:00:00"}