{"id":"https://openalex.org/W7204093385","doi":"https://doi.org/10.48550/arxiv.2608.20794","title":"Knowing but Not Saying: Preventing Factual Access Failures in LLM SFT via Recall-Anchored Distillation","display_name":"Knowing but Not Saying: Preventing Factual Access Failures in LLM SFT via Recall-Anchored Distillation","publication_year":2026,"publication_date":"2026-08-21","ids":{"openalex":"https://openalex.org/W7204093385","doi":"https://doi.org/10.48550/arxiv.2608.20794"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.20794","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.20794","pdf_url":"https://arxiv.org/pdf/2608.20794","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.20794","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5124937525","display_name":"Haodong Chen","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I9842412","display_name":"Nanjing University of Aeronautics and Astronautics","ror":"https://ror.org/01scyh794","country_code":"CN","type":"education","lineage":["https://openalex.org/I9842412"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chen, Haodong","raw_affiliation_strings":["MIIT Key Laboratory of Pattern Analysis and Machine Intelligence , College of Computer Science and Technology , Nanjing University of Aeronautics and Astronautics"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"MIIT Key Laboratory of Pattern Analysis and Machine Intelligence , College of Computer Science and Technology , Nanjing University of Aeronautics and Astronautics","institution_ids":["https://openalex.org/I9842412"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125716457","display_name":"Yadong Wang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I9842412","display_name":"Nanjing University of Aeronautics and Astronautics","ror":"https://ror.org/01scyh794","country_code":"CN","type":"education","lineage":["https://openalex.org/I9842412"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wang, Yadong","raw_affiliation_strings":["MIIT Key Laboratory of Pattern Analysis and Machine Intelligence , College of Computer Science and Technology , Nanjing University of Aeronautics and Astronautics"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"MIIT Key Laboratory of Pattern Analysis and Machine Intelligence , College of Computer Science and Technology , Nanjing University of Aeronautics and Astronautics","institution_ids":["https://openalex.org/I9842412"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5112950235","display_name":"Shengtao Wen","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I9842412","display_name":"Nanjing University of Aeronautics and Astronautics","ror":"https://ror.org/01scyh794","country_code":"CN","type":"education","lineage":["https://openalex.org/I9842412"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wen, Shengtao","raw_affiliation_strings":["MIIT Key Laboratory of Pattern Analysis and Machine Intelligence , College of Computer Science and Technology , Nanjing University of Aeronautics and Astronautics"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"MIIT Key Laboratory of Pattern Analysis and Machine Intelligence , College of Computer Science and Technology , Nanjing University of Aeronautics and Astronautics","institution_ids":["https://openalex.org/I9842412"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124874738","display_name":"Dong Liang Dong Liang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I9842412","display_name":"Nanjing University of Aeronautics and Astronautics","ror":"https://ror.org/01scyh794","country_code":"CN","type":"education","lineage":["https://openalex.org/I9842412"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Liang, Dong","raw_affiliation_strings":["MIIT Key Laboratory of Pattern Analysis and Machine Intelligence , College of Computer Science and Technology , Nanjing University of Aeronautics and Astronautics"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"MIIT Key Laboratory of Pattern Analysis and Machine Intelligence , College of Computer Science and Technology , Nanjing University of Aeronautics and Astronautics","institution_ids":["https://openalex.org/I9842412"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5148611296","display_name":"Xiang Chen","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I9842412","display_name":"Nanjing University of Aeronautics and Astronautics","ror":"https://ror.org/01scyh794","country_code":"CN","type":"education","lineage":["https://openalex.org/I9842412"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chen, Xiang","raw_affiliation_strings":["MIIT Key Laboratory of Pattern Analysis and Machine Intelligence , College of Computer Science and Technology , Nanjing University of Aeronautics and Astronautics"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"MIIT Key Laboratory of Pattern Analysis and Machine Intelligence , College of Computer Science and Technology , Nanjing University of Aeronautics and Astronautics","institution_ids":["https://openalex.org/I9842412"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I9842412"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.5795999765396118,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.5795999765396118,"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":0.3882000148296356,"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":0.006000000052154064,"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/supervised-fine-tuning","display_name":"supervised fine-tuning","score":0.8209999799728394},{"id":"https://openalex.org/keywords/catastrophic-forgetting","display_name":"catastrophic forgetting","score":0.8730000257492065},{"id":"https://openalex.org/keywords/self-distillation","display_name":"self-distillation","score":0.8679999709129333},{"id":"https://openalex.org/keywords/medical-question-answering","display_name":"medical question answering","score":0.6179999709129333},{"id":"https://openalex.org/keywords/domain-adaptation","display_name":"domain adaptation","score":0.4399999976158142},{"id":"https://openalex.org/keywords/open-ended-generation","display_name":"open-ended generation","score":0.4180000126361847}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6686000227928162},{"id":"https://openalex.org/C36503486","wikidata":"https://www.wikidata.org/wiki/Q11235244","display_name":"Domain (mathematical analysis)","level":2,"score":0.4973999857902527},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.48089998960494995},{"id":"https://openalex.org/C164226766","wikidata":"https://www.wikidata.org/wiki/Q7293202","display_name":"Rank (graph theory)","level":2,"score":0.42480000853538513},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3817000091075897},{"id":"https://openalex.org/C42058472","wikidata":"https://www.wikidata.org/wiki/Q810214","display_name":"Base (topology)","level":2,"score":0.311599999666214},{"id":"https://openalex.org/C88006597","wikidata":"https://www.wikidata.org/wiki/Q690117","display_name":"Disk formatting","level":2,"score":0.30640000104904175},{"id":"https://openalex.org/C200601418","wikidata":"https://www.wikidata.org/wiki/Q2193887","display_name":"Reliability engineering","level":1,"score":0.30390000343322754},{"id":"https://openalex.org/C2779679103","wikidata":"https://www.wikidata.org/wiki/Q5251805","display_name":"Degradation (telecommunications)","level":2,"score":0.30309998989105225},{"id":"https://openalex.org/C88626702","wikidata":"https://www.wikidata.org/wiki/Q1128903","display_name":"Continuation","level":2,"score":0.29100000858306885},{"id":"https://openalex.org/C100660578","wikidata":"https://www.wikidata.org/wiki/Q18733","display_name":"Recall","level":2,"score":0.25999999046325684},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.25859999656677246},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.25040000677108765}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.20794","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.20794","pdf_url":"https://arxiv.org/pdf/2608.20794","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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"},{"id":"doi:10.48550/arxiv.2608.20794","doi":"https://doi.org/10.48550/arxiv.2608.20794","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.20794","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2608.20794","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.20794","pdf_url":"https://arxiv.org/pdf/2608.20794","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7204093385.pdf","grobid_xml":"https://content.openalex.org/works/W7204093385.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7204881394","https://openalex.org/W7163223704","https://openalex.org/W7154967367","https://openalex.org/W7119495347","https://openalex.org/W4417308166","https://openalex.org/W7161709916","https://openalex.org/W7221171917","https://openalex.org/W7167580395","https://openalex.org/W7214761566","https://openalex.org/W7148751396"],"abstract_inverted_index":{"Supervised":[0],"fine-tuning":[1],"(SFT)":[2],"can":[3,50],"degrade":[4],"factual":[5,21,43,84,147],"behavior":[6,119],"outside":[7],"the":[8,28,55,122,126,161,173,180,185],"target":[9],"domain.":[10],"This":[11],"degradation":[12,85],"is":[13,184],"often":[14],"described":[15],"as":[16,95],"catastrophic":[17],"forgetting,":[18],"yet":[19],"open-ended":[20],"failures":[22,93],"do":[23],"not":[24],"necessarily":[25],"imply":[26],"that":[27,82,115,179],"underlying":[29],"facts":[30],"have":[31],"been":[32],"erased.":[33],"In":[34],"this":[35,104],"work,":[36],"we":[37,80,106],"identify":[38],"a":[39,111,157],"more":[40],"specific":[41],"phenomenon,":[42],"access":[44],"failure:":[45],"after":[46],"domain":[47],"SFT,":[48],"models":[49],"still":[51],"recognize":[52],"or":[53,145],"rank":[54],"correct":[56],"answer":[57],"under":[58],"constrained":[59],"evaluation,":[60],"while":[61,165],"failing":[62],"to":[63],"produce":[64],"it":[65],"in":[66],"closed-book":[67],"generation.":[68],"Through":[69],"benchmark-level":[70],"comparisons,":[71],"same-fact":[72],"multiple-choice":[73],"and":[74,77,91,99],"generation":[75,118],"probes,":[76],"failure-mode":[78],"analysis,":[79],"show":[81],"SFT-induced":[83],"reflects":[86],"both":[87],"genuine":[88],"wrong-answer":[89],"generations":[90],"expression-level":[92],"such":[94],"verbosity,":[96],"formatting":[97],"mismatch,":[98],"exact-match":[100],"artifacts.":[101],"To":[102],"address":[103],"problem,":[105],"introduce":[107],"Recall-Anchored":[108],"Distillation":[109],"(RAD),":[110],"base-anchored":[112],"self-distillation":[113],"objective":[114],"preserves":[116],"out-of-distribution":[117],"by":[120],"aligning":[121],"adapted":[123],"model":[124],"with":[125,170],"original":[127],"base":[128,186],"model's":[129,187],"soft":[130,188],"continuation":[131],"distribution":[132,189],"on":[133,153,172],"unlabeled":[134],"OOD":[135,141,163,175],"text.":[136],"RAD":[137,155,177],"requires":[138],"no":[139],"gold":[140],"answers,":[142],"external":[143],"judges,":[144],"labeled":[146],"data.":[148],"Across":[149],"three":[150],"backbones":[151],"fine-tuned":[152],"MedMCQA,":[154],"recovers":[156],"consistent":[158],"portion":[159],"of":[160],"lost":[162],"recall":[164],"preserving":[166],"target-domain":[167],"adaptation.":[168],"Compared":[169],"replay":[171],"same":[174],"text,":[176],"shows":[178],"key":[181],"preservation":[182],"signal":[183],"rather":[190],"than":[191],"additional":[192],"text":[193],"exposure":[194],"alone.":[195]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-08-25T00:00:00"}