{"id":"https://openalex.org/W7128717306","doi":"https://doi.org/10.48550/arxiv.2602.10352","title":"Learning Self-Interpretation from Interpretability Artifacts: Training Lightweight Adapters on Vector-Label Pairs","display_name":"Learning Self-Interpretation from Interpretability Artifacts: Training Lightweight Adapters on Vector-Label Pairs","publication_year":2026,"publication_date":"2026-02-10","ids":{"openalex":"https://openalex.org/W7128717306","doi":"https://doi.org/10.48550/arxiv.2602.10352"},"language":null,"primary_location":{"id":"pmh:doi:10.48550/arxiv.2602.10352","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"Article"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":null,"any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5034943857","display_name":"Keenan Pepper","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pepper, Keenan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125383458","display_name":"Alex McKenzie","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"McKenzie, Alex","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Pop, Florin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Pop, Florin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125423495","display_name":"Stijn Servaes","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Servaes, Stijn","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5121317525","display_name":"Martin Leitgab","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Leitgab, Martin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5094342551","display_name":"Mike Vaiana","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Vaiana, Mike","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5099801492","display_name":"Judd Rosenblatt","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Rosenblatt, Judd","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125698547","display_name":"Michael S. A. Graziano","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Graziano, Michael S. A.","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5099801493","display_name":"Diogo Schwerz de Lucena","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"de Lucena, Diogo","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":9,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"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":0.8077999949455261,"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":0.8077999949455261,"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":0.05290000140666962,"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":0.018699999898672104,"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":0.9351000189781189},{"id":"https://openalex.org/keywords/autoencoder","display_name":"Autoencoder","score":0.6388999819755554},{"id":"https://openalex.org/keywords/leverage","display_name":"Leverage (statistics)","score":0.4738999903202057},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.4693000018596649},{"id":"https://openalex.org/keywords/adapter","display_name":"Adapter (computing)","score":0.46059998869895935},{"id":"https://openalex.org/keywords/axiom","display_name":"Axiom","score":0.400299996137619},{"id":"https://openalex.org/keywords/discriminative-model","display_name":"Discriminative model","score":0.38100001215934753},{"id":"https://openalex.org/keywords/affine-transformation","display_name":"Affine transformation","score":0.38029998540878296}],"concepts":[{"id":"https://openalex.org/C2781067378","wikidata":"https://www.wikidata.org/wiki/Q17027399","display_name":"Interpretability","level":2,"score":0.9351000189781189},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7139000296592712},{"id":"https://openalex.org/C101738243","wikidata":"https://www.wikidata.org/wiki/Q786435","display_name":"Autoencoder","level":3,"score":0.6388999819755554},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6176999807357788},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4878999888896942},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.4738999903202057},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.4693000018596649},{"id":"https://openalex.org/C177284502","wikidata":"https://www.wikidata.org/wiki/Q1005390","display_name":"Adapter (computing)","level":2,"score":0.46059998869895935},{"id":"https://openalex.org/C167729594","wikidata":"https://www.wikidata.org/wiki/Q17736","display_name":"Axiom","level":2,"score":0.400299996137619},{"id":"https://openalex.org/C97931131","wikidata":"https://www.wikidata.org/wiki/Q5282087","display_name":"Discriminative model","level":2,"score":0.38100001215934753},{"id":"https://openalex.org/C92757383","wikidata":"https://www.wikidata.org/wiki/Q382497","display_name":"Affine transformation","level":2,"score":0.38029998540878296},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.3707999885082245},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.36320000886917114},{"id":"https://openalex.org/C100776233","wikidata":"https://www.wikidata.org/wiki/Q2532492","display_name":"Bridge (graph theory)","level":2,"score":0.3594000041484833},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.319599986076355},{"id":"https://openalex.org/C8642999","wikidata":"https://www.wikidata.org/wiki/Q4171168","display_name":"Hyperparameter","level":2,"score":0.29660001397132874},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.2883000075817108},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.2858999967575073},{"id":"https://openalex.org/C83665646","wikidata":"https://www.wikidata.org/wiki/Q42139305","display_name":"Feature vector","level":2,"score":0.28029999136924744},{"id":"https://openalex.org/C170858558","wikidata":"https://www.wikidata.org/wiki/Q1394144","display_name":"Automatic summarization","level":2,"score":0.27250000834465027},{"id":"https://openalex.org/C2776809875","wikidata":"https://www.wikidata.org/wiki/Q1375963","display_name":"Converse","level":2,"score":0.2721000015735626},{"id":"https://openalex.org/C55166926","wikidata":"https://www.wikidata.org/wiki/Q2892946","display_name":"Oracle","level":2,"score":0.2660999894142151}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:doi:10.48550/arxiv.2602.10352","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"Article"},{"id":"doi:10.48550/arxiv.2602.10352","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.10352","pdf_url":null,"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"article"}],"best_oa_location":{"id":"pmh:doi:10.48550/arxiv.2602.10352","is_oa":true,"landing_page_url":null,"pdf_url":null,"source":{"id":"https://openalex.org/S4406922384","display_name":"Open MIND","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"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":"Article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Self-interpretation":[0],"methods":[1],"prompt":[2,92],"language":[3],"models":[4],"to":[5,15,135],"describe":[6],"their":[7],"own":[8],"internal":[9],"states,":[10],"but":[11],"remain":[12],"unreliable":[13],"due":[14],"hyperparameter":[16],"sensitivity.":[17],"We":[18],"show":[19],"that":[20,57,88,141],"training":[21,60],"lightweight":[22],"adapters":[23,51,112],"on":[24],"interpretability":[25],"artifacts,":[26],"while":[27],"keeping":[28],"the":[29,59,148],"LM":[30],"entirely":[31],"frozen,":[32],"yields":[33],"reliable":[34],"self-interpretation":[35,128,142],"across":[36],"tasks":[37],"and":[38,81,110],"model":[39,121,149],"families.":[40],"A":[41],"scalar":[42],"affine":[43],"adapter":[44],"with":[45,73,144],"just":[46],"$d_\\text{model}+1$":[47],"parameters":[48],"suffices:":[49],"trained":[50],"generate":[52],"sparse":[53],"autoencoder":[54],"feature":[55],"labels":[56,61],"outperform":[58],"themselves":[62],"(70%":[63],"vs":[64],"50%":[65],"generation":[66],"scoring":[67],"at":[68],"70B":[69],"scale),":[70],"identify":[71],"topics":[72],"94%":[74],"recall@1":[75],"versus":[76],"1%":[77],"for":[78,106,120],"untrained":[79],"baselines,":[80],"decode":[82],"bridge":[83],"entities":[84],"in":[85,90],"multi-hop":[86],"reasoning":[87,97],"appear":[89],"neither":[91],"nor":[93],"response,":[94],"surfacing":[95],"implicit":[96],"without":[98,146],"chain-of-thought.":[99],"The":[100],"learned":[101],"bias":[102],"vector":[103],"alone":[104],"accounts":[105],"85%":[107],"of":[108],"improvement,":[109],"simpler":[111],"generalize":[113],"better":[114],"than":[115],"more":[116],"expressive":[117],"alternatives.":[118],"Controlling":[119],"knowledge":[122],"via":[123],"prompted":[124],"descriptions,":[125],"we":[126],"find":[127],"gains":[129,132],"outpace":[130],"capability":[131],"from":[133],"7B":[134],"72B":[136],"parameters.":[137],"Our":[138],"results":[139],"demonstrate":[140],"improves":[143],"scale,":[145],"modifying":[147],"being":[150],"interpreted.":[151]},"counts_by_year":[],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2026-02-13T00:00:00"}