{"id":"https://openalex.org/W4409503125","doi":"https://doi.org/10.1162/tacl_a_00741","title":"Self-Rationalization in the Wild: A Large-scale Out-of-Distribution Evaluation on NLI-related tasks","display_name":"Self-Rationalization in the Wild: A Large-scale Out-of-Distribution Evaluation on NLI-related tasks","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4409503125","doi":"https://doi.org/10.1162/tacl_a_00741"},"language":"en","primary_location":{"id":"doi:10.1162/tacl_a_00741","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl_a_00741","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_00741","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5009587198","display_name":"Jing Yang","orcid":"https://orcid.org/0000-0002-5918-2991"},"institutions":[{"id":"https://openalex.org/I181391015","display_name":"Universidade Estadual de Campinas (UNICAMP)","ror":"https://ror.org/04wffgt70","country_code":"BR","type":"education","lineage":["https://openalex.org/I181391015"]}],"countries":["BR"],"is_corresponding":true,"raw_author_name":"Jing Yang","raw_affiliation_strings":["Artificial Intelligence Lab., Recod.ai, Institute of Computing, University of Campinas, Brazil. jing.yang@ic.unicamp.br"],"affiliations":[{"raw_affiliation_string":"Artificial Intelligence Lab., Recod.ai, Institute of Computing, University of Campinas, Brazil. jing.yang@ic.unicamp.br","institution_ids":["https://openalex.org/I181391015"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5021242117","display_name":"Max Glockner","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":"Max Glockner","raw_affiliation_strings":["UKP Lab, Department of Computer Science, Technical University of Darmstadt, Germany. max.glockner@tu-darmstadt.de"],"affiliations":[{"raw_affiliation_string":"UKP Lab, Department of Computer Science, Technical University of Darmstadt, Germany. max.glockner@tu-darmstadt.de","institution_ids":["https://openalex.org/I31512782"]}]},{"author_position":"middle","author":{"id":null,"display_name":"Anderson Rocha","orcid":null},"institutions":[{"id":"https://openalex.org/I181391015","display_name":"Universidade Estadual de Campinas (UNICAMP)","ror":"https://ror.org/04wffgt70","country_code":"BR","type":"education","lineage":["https://openalex.org/I181391015"]}],"countries":["BR"],"is_corresponding":false,"raw_author_name":"Anderson Rocha","raw_affiliation_strings":["Artificial Intelligence Lab., Recod.ai, Institute of Computing, University of Campinas, Brazil. anderson.rocha@unicamp.br"],"affiliations":[{"raw_affiliation_string":"Artificial Intelligence Lab., Recod.ai, Institute of Computing, University of Campinas, Brazil. anderson.rocha@unicamp.br","institution_ids":["https://openalex.org/I181391015"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5027450194","display_name":"Iryna Gurevych","orcid":"https://orcid.org/0000-0003-2187-7621"},"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":["UKP Lab, Department of Computer Science, Technical University of Darmstadt, Germany. iryna.gurevych@tu-darmstadt.de"],"affiliations":[{"raw_affiliation_string":"UKP Lab, Department of Computer Science, Technical University of Darmstadt, Germany. iryna.gurevych@tu-darmstadt.de","institution_ids":["https://openalex.org/I31512782"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":4,"corresponding_author_ids":["https://openalex.org/A5009587198"],"corresponding_institution_ids":["https://openalex.org/I181391015"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.03460268,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":"13","issue":null,"first_page":"314","last_page":"342"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9994000196456909,"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/T10028","display_name":"Topic Modeling","score":0.9994000196456909,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.9843000173568726,"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/T12128","display_name":"AI in Service Interactions","score":0.9796000123023987,"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/rationalization","display_name":"Rationalization (economics)","score":0.8050472736358643},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7995791435241699},{"id":"https://openalex.org/keywords/scale","display_name":"Scale (ratio)","score":0.5173079967498779},{"id":"https://openalex.org/keywords/data-science","display_name":"Data science","score":0.33654141426086426},{"id":"https://openalex.org/keywords/epistemology","display_name":"Epistemology","score":0.08731529116630554},{"id":"https://openalex.org/keywords/physics","display_name":"Physics","score":0.05629405379295349}],"concepts":[{"id":"https://openalex.org/C52438962","wikidata":"https://www.wikidata.org/wiki/Q1555139","display_name":"Rationalization (economics)","level":2,"score":0.8050472736358643},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7995791435241699},{"id":"https://openalex.org/C2778755073","wikidata":"https://www.wikidata.org/wiki/Q10858537","display_name":"Scale (ratio)","level":2,"score":0.5173079967498779},{"id":"https://openalex.org/C2522767166","wikidata":"https://www.wikidata.org/wiki/Q2374463","display_name":"Data science","level":1,"score":0.33654141426086426},{"id":"https://openalex.org/C111472728","wikidata":"https://www.wikidata.org/wiki/Q9471","display_name":"Epistemology","level":1,"score":0.08731529116630554},{"id":"https://openalex.org/C121332964","wikidata":"https://www.wikidata.org/wiki/Q413","display_name":"Physics","level":0,"score":0.05629405379295349},{"id":"https://openalex.org/C138885662","wikidata":"https://www.wikidata.org/wiki/Q5891","display_name":"Philosophy","level":0,"score":0.0},{"id":"https://openalex.org/C62520636","wikidata":"https://www.wikidata.org/wiki/Q944","display_name":"Quantum mechanics","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.1162/tacl_a_00741","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl_a_00741","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:154938","is_oa":false,"landing_page_url":"http://tubiblio.ulb.tu-darmstadt.de/view/person/Yang=3AJing=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":"NonPeerReviewed"}],"best_oa_location":{"id":"doi:10.1162/tacl_a_00741","is_oa":true,"landing_page_url":"https://doi.org/10.1162/tacl_a_00741","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":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":87,"referenced_works":["https://openalex.org/W398859631","https://openalex.org/W569478347","https://openalex.org/W1537768638","https://openalex.org/W1840435438","https://openalex.org/W2515796212","https://openalex.org/W2548036585","https://openalex.org/W2771976988","https://openalex.org/W2889468083","https://openalex.org/W2890771450","https://openalex.org/W2923014074","https://openalex.org/W2941666437","https://openalex.org/W2947337775","https://openalex.org/W2951286828","https://openalex.org/W2951936329","https://openalex.org/W2962843521","https://openalex.org/W2963961878","https://openalex.org/W2988245244","https://openalex.org/W3034188538","https://openalex.org/W3034383590","https://openalex.org/W3034937117","https://openalex.org/W3034999214","https://openalex.org/W3035139434","https://openalex.org/W3035503910","https://openalex.org/W3098987177","https://openalex.org/W3099977667","https://openalex.org/W3103649165","https://openalex.org/W3106234277","https://openalex.org/W3168867926","https://openalex.org/W3170180819","https://openalex.org/W3172045361","https://openalex.org/W3174481471","https://openalex.org/W3175591618","https://openalex.org/W3176640961","https://openalex.org/W3183138634","https://openalex.org/W3206132631","https://openalex.org/W3212191244","https://openalex.org/W3212748247","https://openalex.org/W4221143046","https://openalex.org/W4224088803","https://openalex.org/W4225468040","https://openalex.org/W4281493163","https://openalex.org/W4285240908","https://openalex.org/W4287854450","https://openalex.org/W4288089799","https://openalex.org/W4297633153","https://openalex.org/W4304697829","https://openalex.org/W4363671827","https://openalex.org/W4365799947","https://openalex.org/W4384918448","https://openalex.org/W4385569970","https://openalex.org/W4385571815","https://openalex.org/W4385572910","https://openalex.org/W4385573270","https://openalex.org/W4385573608","https://openalex.org/W4386080541","https://openalex.org/W4387323837","https://openalex.org/W4389518953","https://openalex.org/W4393027574","https://openalex.org/W4399116034","https://openalex.org/W4400104857","https://openalex.org/W4401042762","https://openalex.org/W4401042888","https://openalex.org/W4401043276","https://openalex.org/W4402670492","https://openalex.org/W4402671025","https://openalex.org/W4402671236","https://openalex.org/W4402671286","https://openalex.org/W4405883687","https://openalex.org/W6691303741","https://openalex.org/W6745573522","https://openalex.org/W6754655096","https://openalex.org/W6761205521","https://openalex.org/W6769627184","https://openalex.org/W6786063503","https://openalex.org/W6809646742","https://openalex.org/W6810464575","https://openalex.org/W6838648451","https://openalex.org/W6852252866","https://openalex.org/W6854866820","https://openalex.org/W6856154222","https://openalex.org/W6856652384","https://openalex.org/W6857253974","https://openalex.org/W6858460891","https://openalex.org/W6861315699","https://openalex.org/W6862618468","https://openalex.org/W6876225394","https://openalex.org/W6947351026"],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2980745308","https://openalex.org/W2380441323","https://openalex.org/W2006855656","https://openalex.org/W1990734973","https://openalex.org/W434944249","https://openalex.org/W4234347025","https://openalex.org/W3015389025"],"abstract_inverted_index":{"Abstract":[0],"Free-text":[1],"explanations":[2],"are":[3,69],"expressive":[4],"and":[5,38,47,50,63,84,104,113,169],"easy":[6],"to":[7,19,31,155,178],"understand,":[8],"but":[9],"many":[10],"datasets":[11,35,75],"lack":[12],"annotated":[13,144],"explanation":[14,34,93,151],"data,":[15],"making":[16],"it":[17],"challenging":[18],"train":[20],"models":[21,49,68,103,148,171],"for":[22,36,149],"explainable":[23],"predictions.":[24],"To":[25],"address":[26],"this,":[27],"we":[28,95],"investigate":[29],"how":[30],"use":[32],"existing":[33],"self-rationalization":[37],"evaluate":[39],"models\u2019":[40],"out-of-distribution":[41],"(OOD)":[42],"performance.":[43],"We":[44],"fine-tune":[45],"T5-Large":[46],"OLMo-7B":[48],"assess":[51],"the":[52,58,91,109,123],"impact":[53,165],"of":[54,60],"fine-tuning":[55,61,159],"data":[56,160],"quality,":[57],"number":[59],"samples,":[62],"few-shot":[64],"selection":[65,157],"methods.":[66],"The":[67],"evaluated":[70],"on":[71,100,166],"19":[72],"diverse":[73],"OOD":[74,150,167],"across":[76],"three":[77,114],"tasks:":[78],"natural":[79],"language":[80],"inference":[81],"(NLI),":[82],"fact-checking,":[83],"hallucination":[85],"detection":[86],"in":[87,135],"abstractive":[88],"summarization.":[89],"For":[90],"generated":[92],"evaluation,":[94],"conduct":[96],"a":[97,163],"human":[98,130],"study":[99,105],"13":[101],"selected":[102],"its":[106,133],"correlation":[107],"with":[108,129,172],"Acceptability":[110,124,186],"score":[111,125],"(T5-11B)":[112],"other":[115],"LLM-based":[116],"reference-free":[117],"metrics.":[118],"Human":[119],"evaluation":[120],"shows":[121],"that":[122],"correlates":[126],"most":[127],"strongly":[128],"judgments,":[131],"demonstrating":[132],"effectiveness":[134],"evaluating":[136],"free-text":[137],"explanations.":[138],"Our":[139],"findings":[140],"reveal:":[141],"1)":[142],"few":[143],"examples":[145],"effectively":[146],"adapt":[147],"generation;":[152],"2)":[153],"compared":[154],"sample":[156],"strategies,":[158],"source":[161],"has":[162],"larger":[164],"performance;":[168],"3)":[170],"higher":[173,185],"label":[174],"prediction":[175],"accuracy":[176],"tend":[177],"produce":[179],"better":[180],"explanations,":[181],"as":[182],"reflected":[183],"by":[184],"scores.1":[187]},"counts_by_year":[],"updated_date":"2026-03-25T23:56:10.502304","created_date":"2025-10-10T00:00:00"}