{"id":"https://openalex.org/W7150845224","doi":"https://doi.org/10.1162/tacl.a.613","title":"Aligned Probing: Relating Toxic Behavior and Model Internals","display_name":"Aligned Probing: Relating Toxic Behavior and Model Internals","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7150845224","doi":"https://doi.org/10.1162/tacl.a.613"},"language":"en","primary_location":{"id":"doi:10.1162/tacl.a.613","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl.a.613","pdf_url":null,"source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"},"type":"article","indexed_in":["crossref","doaj"],"open_access":{"is_oa":true,"oa_status":"diamond","oa_url":"https://doi.org/10.1162/tacl.a.613","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5085003328","display_name":"Andreas Waldis","orcid":"https://orcid.org/0000-0002-2772-5701"},"institutions":[{"id":"https://openalex.org/I31512782","display_name":"Technische Universit\u00e4t Darmstadt","ror":"https://ror.org/05n911h24","country_code":"DE","type":"education","lineage":["https://openalex.org/I31512782"]},{"id":"https://openalex.org/I81007117","display_name":"Lucerne University of Applied Sciences and Arts","ror":"https://ror.org/04nd0xd48","country_code":"CH","type":"education","lineage":["https://openalex.org/I81007117"]}],"countries":["CH","DE"],"is_corresponding":true,"raw_author_name":"Andreas Waldis","raw_affiliation_strings":["Information Systems Research Lab, Lucerne University of Applied Sciences and Arts, Switzerland","Ubiquitous Knowledge Processing Lab (UKP Lab), Technical University of Darmstadt, Germany"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Information Systems Research Lab, Lucerne University of Applied Sciences and Arts, Switzerland","institution_ids":["https://openalex.org/I81007117"]},{"raw_affiliation_string":"Ubiquitous Knowledge Processing Lab (UKP Lab), Technical University of Darmstadt, Germany","institution_ids":["https://openalex.org/I31512782"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5013587924","display_name":"Vagrant Gautam","orcid":"https://orcid.org/0000-0002-7263-8578"},"institutions":[{"id":"https://openalex.org/I91712215","display_name":"Saarland University","ror":"https://ror.org/01jdpyv68","country_code":"DE","type":"education","lineage":["https://openalex.org/I91712215"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Vagrant Gautam","raw_affiliation_strings":["Spoken Language Systems, Saarland University, Germany"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Spoken Language Systems, Saarland University, Germany","institution_ids":["https://openalex.org/I91712215"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5022688200","display_name":"Anne Lauscher","orcid":"https://orcid.org/0000-0001-8590-9827"},"institutions":[{"id":"https://openalex.org/I159176309","display_name":"Universit\u00e4t Hamburg","ror":"https://ror.org/00g30e956","country_code":"DE","type":"education","lineage":["https://openalex.org/I159176309"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Anne Lauscher","raw_affiliation_strings":["Data Science Group, University of Hamburg, Germany"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Data Science Group, University of Hamburg, Germany","institution_ids":["https://openalex.org/I159176309"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5008875255","display_name":"Dietrich Klakow","orcid":"https://orcid.org/0000-0002-4147-9690"},"institutions":[{"id":"https://openalex.org/I91712215","display_name":"Saarland University","ror":"https://ror.org/01jdpyv68","country_code":"DE","type":"education","lineage":["https://openalex.org/I91712215"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Dietrich Klakow","raw_affiliation_strings":["Spoken Language Systems, Saarland University, Germany"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Spoken Language Systems, Saarland University, Germany","institution_ids":["https://openalex.org/I91712215"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5133037463","display_name":"Iryna Gurevych","orcid":null},"institutions":[{"id":"https://openalex.org/I31512782","display_name":"Technische Universit\u00e4t Darmstadt","ror":"https://ror.org/05n911h24","country_code":"DE","type":"education","lineage":["https://openalex.org/I31512782"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Iryna Gurevych","raw_affiliation_strings":["Ubiquitous Knowledge Processing Lab (UKP Lab), Technical University of Darmstadt, Germany"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Ubiquitous Knowledge Processing Lab (UKP Lab), Technical University of Darmstadt, Germany","institution_ids":["https://openalex.org/I31512782"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5085003328"],"corresponding_institution_ids":["https://openalex.org/I31512782","https://openalex.org/I81007117"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.39295412,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"14","issue":null,"first_page":"271","last_page":"291"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12262","display_name":"Hate Speech and Cyberbullying Detection","score":0.7907000184059143,"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/T12262","display_name":"Hate Speech and Cyberbullying Detection","score":0.7907000184059143,"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/T11147","display_name":"Misinformation and Its Impacts","score":0.0430000014603138,"subfield":{"id":"https://openalex.org/subfields/3312","display_name":"Sociology and Political Science"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.035100001841783524,"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":0.9502000212669373},{"id":"https://openalex.org/keywords/bridging","display_name":"Bridging (networking)","score":0.680899977684021},{"id":"https://openalex.org/keywords/offensive","display_name":"Offensive","score":0.6344000101089478},{"id":"https://openalex.org/keywords/context","display_name":"Context (archaeology)","score":0.6121000051498413},{"id":"https://openalex.org/keywords/encode","display_name":"ENCODE","score":0.5378999710083008},{"id":"https://openalex.org/keywords/context-model","display_name":"Context model","score":0.41990000009536743}],"concepts":[{"id":"https://openalex.org/C2781067378","wikidata":"https://www.wikidata.org/wiki/Q17027399","display_name":"Interpretability","level":2,"score":0.9502000212669373},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8066999912261963},{"id":"https://openalex.org/C174348530","wikidata":"https://www.wikidata.org/wiki/Q188635","display_name":"Bridging (networking)","level":2,"score":0.680899977684021},{"id":"https://openalex.org/C176856949","wikidata":"https://www.wikidata.org/wiki/Q2001676","display_name":"Offensive","level":2,"score":0.6344000101089478},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.6121000051498413},{"id":"https://openalex.org/C66746571","wikidata":"https://www.wikidata.org/wiki/Q1134833","display_name":"ENCODE","level":3,"score":0.5378999710083008},{"id":"https://openalex.org/C183322885","wikidata":"https://www.wikidata.org/wiki/Q17007702","display_name":"Context model","level":3,"score":0.41990000009536743},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4049000144004822},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.39070001244544983},{"id":"https://openalex.org/C145912823","wikidata":"https://www.wikidata.org/wiki/Q113558","display_name":"Dynamics (music)","level":2,"score":0.3375000059604645},{"id":"https://openalex.org/C2776800370","wikidata":"https://www.wikidata.org/wiki/Q5172864","display_name":"Correlative","level":2,"score":0.30730000138282776},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.29899999499320984},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.290800005197525},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.2872999906539917},{"id":"https://openalex.org/C28427503","wikidata":"https://www.wikidata.org/wiki/Q13580300","display_name":"Internal model","level":3,"score":0.2842999994754791},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.27869999408721924}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1162/tacl.a.613","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl.a.613","pdf_url":null,"source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"},{"id":"pmh:oai:tubiblio.ulb.tu-darmstadt.de:160017","is_oa":false,"landing_page_url":"http://tubiblio.ulb.tu-darmstadt.de/view/person/Waldis=3AAndreas=3A=3A.html>","pdf_url":null,"source":{"id":"https://openalex.org/S4377196390","display_name":"TUbilio (Technical University of Darmstadt)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":"https://openalex.org/I31512782","host_organization_name":"Technische Universit\u00e4t Darmstadt","host_organization_lineage":["https://openalex.org/I31512782"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"acceptedVersion","is_accepted":true,"is_published":false,"raw_source_name":null,"raw_type":"Artikel"}],"best_oa_location":{"id":"doi:10.1162/tacl.a.613","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl.a.613","pdf_url":null,"source":{"id":"https://openalex.org/S2729999759","display_name":"Transactions of the Association for Computational Linguistics","issn_l":"2307-387X","issn":["2307-387X"],"is_oa":true,"is_in_doaj":true,"is_core":true,"host_organization":"https://openalex.org/P4310320244","host_organization_name":"Association for Computational Linguistics","host_organization_lineage":["https://openalex.org/P4310320244"],"host_organization_lineage_names":["Association for Computational Linguistics"],"type":"journal"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Transactions of the Association for Computational Linguistics","raw_type":"journal-article"},"sustainable_development_goals":[{"score":0.7209972143173218,"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":39,"referenced_works":["https://openalex.org/W2563826943","https://openalex.org/W2941666437","https://openalex.org/W2946417913","https://openalex.org/W2963943967","https://openalex.org/W2970862333","https://openalex.org/W2996728628","https://openalex.org/W3092448486","https://openalex.org/W3100355250","https://openalex.org/W3104235057","https://openalex.org/W3155744586","https://openalex.org/W3159900299","https://openalex.org/W3173465197","https://openalex.org/W3202070718","https://openalex.org/W3212327893","https://openalex.org/W4285127912","https://openalex.org/W4285210452","https://openalex.org/W4385567161","https://openalex.org/W4386566962","https://openalex.org/W4386576626","https://openalex.org/W4388691863","https://openalex.org/W4389518756","https://openalex.org/W4389519274","https://openalex.org/W4389520710","https://openalex.org/W4389524162","https://openalex.org/W4390091989","https://openalex.org/W4399061743","https://openalex.org/W4399528455","https://openalex.org/W4401042100","https://openalex.org/W4401042685","https://openalex.org/W4401208776","https://openalex.org/W4402671236","https://openalex.org/W4402672079","https://openalex.org/W4404781371","https://openalex.org/W4404782953","https://openalex.org/W4404784049","https://openalex.org/W4405205232","https://openalex.org/W4409362072","https://openalex.org/W4414485088","https://openalex.org/W7126404551"],"related_works":[],"abstract_inverted_index":{"Abstract":[0],"Warning:":[1],"This":[2],"paper":[3],"contains":[4],"offensive":[5],"text.":[6],"We":[7,102],"introduce":[8],"aligned":[9,139],"probing,":[10],"a":[11,149],"novel":[12],"interpretability":[13],"framework":[14],"that":[15,58,88],"aligns":[16],"the":[17,52,64,99,105,135,159],"behavior":[18,111],"of":[19,67,107,138,153,161],"language":[20],"models":[21],"(LMs),":[22],"based":[23],"on":[24,77],"their":[25,28],"outputs,":[26,71],"and":[27,41,46,69,85,112,131,157],"internal":[29,47],"representations":[30],"(internals).":[31],"Using":[32],"this":[33],"framework,":[34],"we":[35],"examine":[36],"over":[37],"20":[38],"OLMo,":[39],"Llama,":[40],"Mistral":[42],"models,":[43],"bridging":[44],"behavioral":[45],"perspectives":[48],"for":[49,51],"toxicity":[50,65],"first":[53],"time.":[54],"Our":[55,145],"results":[56],"show":[57],"LMs":[59,80],"strongly":[60,95],"encode":[61],"information":[62,97],"about":[63,98],"level":[66],"inputs":[68],"subsequent":[70],"particularly":[72],"in":[73],"lower":[74],"layers.":[75],"Focusing":[76],"how":[78],"unique":[79,116],"differ":[81],"offers":[82],"both":[83,155],"correlative":[84],"causal":[86],"evidence":[87],"they":[89],"generate":[90],"less":[91],"toxic":[92],"output":[93],"when":[94],"encoding":[96],"input":[100],"toxicity.":[101,162],"also":[103],"highlight":[104],"heterogeneity":[106],"toxicity,":[108],"as":[109,119],"model":[110,129],"internals":[113],"vary":[114],"across":[115],"attributes":[117],"such":[118],"Threat.":[120],"Finally,":[121],"four":[122],"case":[123],"studies":[124],"analyzing":[125],"detoxification,":[126],"multi-prompt":[127],"evaluations,":[128],"quantization,":[130],"pre-training":[132],"dynamics":[133],"underline":[134],"practical":[136],"impact":[137],"probing":[140],"with":[141],"further":[142],"concrete":[143],"insights.":[144],"findings":[146],"contribute":[147],"to":[148],"more":[150],"holistic":[151],"understanding":[152],"LMs,":[154],"within":[156],"beyond":[158],"context":[160],"alignedprobing.github.io":[163]},"counts_by_year":[],"updated_date":"2026-07-31T08:31:51.225901","created_date":"2026-04-07T00:00:00"}