{"id":"https://openalex.org/W7225140147","doi":"https://doi.org/10.48550/arxiv.2611.02746","title":"Replay for LLM Continual Learning with Multi-Teacher On-Policy Distillation","display_name":"Replay for LLM Continual Learning with Multi-Teacher On-Policy Distillation","publication_year":2026,"publication_date":"2026-11-02","ids":{"openalex":"https://openalex.org/W7225140147","doi":"https://doi.org/10.48550/arxiv.2611.02746"},"language":"en","primary_location":{"id":"doi:10.48550/arxiv.2611.02746","doi":"https://doi.org/10.48550/arxiv.2611.02746","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2611.02746","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2611.02746","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5160512649","display_name":"Yiqiao Qiu","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Qiu, Yiqiao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":["https://openalex.org/A5160512649"],"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.9848999977111816,"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":0.9848999977111816,"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":0.007699999958276749,"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":0.002899999963119626,"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/replay","display_name":"replay","score":0.5370000004768372},{"id":"https://openalex.org/keywords/attention-alignment","display_name":"attention alignment","score":0.7269999980926514},{"id":"https://openalex.org/keywords/value-alignment","display_name":"value alignment","score":0.7390000224113464},{"id":"https://openalex.org/keywords/safety-benchmark","display_name":"safety benchmark","score":0.5270000100135803},{"id":"https://openalex.org/keywords/multistage-distillation","display_name":"multistage distillation","score":0.4749999940395355}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7348999977111816},{"id":"https://openalex.org/C28719098","wikidata":"https://www.wikidata.org/wiki/Q44946","display_name":"Point (geometry)","level":2,"score":0.4056999981403351},{"id":"https://openalex.org/C204030448","wikidata":"https://www.wikidata.org/wiki/Q101017","display_name":"Distillation","level":2,"score":0.36309999227523804},{"id":"https://openalex.org/C36503486","wikidata":"https://www.wikidata.org/wiki/Q11235244","display_name":"Domain (mathematical analysis)","level":2,"score":0.319599986076355},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.31839999556541443},{"id":"https://openalex.org/C11560541","wikidata":"https://www.wikidata.org/wiki/Q1756025","display_name":"Replay attack","level":3,"score":0.2808000147342682},{"id":"https://openalex.org/C61445026","wikidata":"https://www.wikidata.org/wiki/Q217608","display_name":"Fixed point","level":2,"score":0.28049999475479126},{"id":"https://openalex.org/C111640148","wikidata":"https://www.wikidata.org/wiki/Q847349","display_name":"Rubric","level":2,"score":0.2655999958515167},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.2565000057220459},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.24979999661445618}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2611.02746","doi":"https://doi.org/10.48550/arxiv.2611.02746","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2611.02746","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2611.02746","doi":"https://doi.org/10.48550/arxiv.2611.02746","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2611.02746","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"display_name":"Quality education","id":"https://metadata.un.org/sdg/4","score":0.6164000034332275}],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7214796688","https://openalex.org/W7167617612","https://openalex.org/W7162543099","https://openalex.org/W7213995943","https://openalex.org/W7214759893","https://openalex.org/W7220430871","https://openalex.org/W7220978264","https://openalex.org/W7203946480","https://openalex.org/W7204475111","https://openalex.org/W7223099873"],"abstract_inverted_index":{"Multi-teacher":[0],"on-policy":[1],"distillation":[2,247],"(MOPD)":[3],"distills":[4],"several":[5],"domain":[6,59],"experts":[7,49,64,153,219],"into":[8],"one":[9,44,221],"student":[10,142,239],"and":[11,47,85,124,201,212,248],"is":[12,60,119,172,215],"used":[13],"in":[14,65],"the":[15,36,48,63,73,104,128,141,144,176,188,237,243],"post-training":[16],"of":[17,50,72,98,106,135,143,162,169,195,207,220,242],"recent":[18,77],"large":[19],"language":[20],"models.":[21],"It":[22],"trains":[23],"on":[24,225,261],"all":[25,94,266],"domains":[26,52,137,178],"at":[27,35],"once,":[28],"so":[29,150],"every":[30,83,132],"expert":[31],"must":[32],"be":[33,54],"available":[34],"same":[37],"time.":[38],"In":[39],"practice,":[40],"capabilities":[41],"often":[42],"arrive":[43],"after":[45,82],"another,":[46],"earlier":[51,74,122,136,152,177,182,257],"may":[53],"gone":[55],"when":[56],"a":[57,80,88,170,252],"new":[58],"learned.":[61],"Distilling":[62],"sequence":[66],"then":[67,101],"suffers":[68],"from":[69,109,121,217,265],"catastrophic":[70],"forgetting":[71],"domains.":[75],"Some":[76],"pipelines":[78],"store":[79],"snapshot":[81],"stage":[84,91,146,155,171],"end":[86],"with":[87,103,140,263],"multi-teacher":[89],"recovery":[90,264],"taught":[92],"by":[93,117],"snapshots.":[95,268],"The":[96,166],"number":[97,105],"stored":[99,267],"models":[100],"grows":[102],"stages.":[107],"Different":[108],"these":[110],"pipelines,":[111],"this":[112],"paper":[113],"formulates":[114],"continual":[115],"MOPD":[116],"what":[118],"kept":[120],"stages":[123],"studies":[125],"replay":[126,163,167,235,254],"toward":[127,236],"previous":[129,145,238],"student.":[130,223],"Inside":[131],"stage,":[133],"prompts":[134],"are":[138,157,164],"replayed":[139],"as":[147],"their":[148],"teacher,":[149],"neither":[151],"nor":[154],"snapshots":[156],"kept.":[158],"Two":[159],"design":[160],"choices":[161],"explored.":[165],"budget":[168],"either":[173,186],"shared":[174],"among":[175],"or":[179,192],"fixed":[180,253],"per":[181,256],"domain.":[183],"Replay":[184],"steps":[185],"distill":[187],"token":[189],"distribution":[190],"alone":[191],"add":[193],"alignment":[194],"hidden":[196],"states,":[197],"region-pooled":[198],"self-attention":[199],"scores":[200],"value":[202],"relations.":[203],"A":[204],"five-domain":[205],"benchmark":[206],"logic,":[208],"math,":[209],"code,":[210],"safety":[211],"instruction":[213],"following":[214],"built":[216],"public":[218],"1.5B":[222],"Experiments":[224],"its":[226],"14":[227],"benchmarks":[228],"show":[229],"that,":[230],"without":[231],"storing":[232],"any":[233],"snapshot,":[234],"closes":[240],"most":[241],"gap":[244],"between":[245],"sequential":[246],"joint":[249],"MOPD.":[250],"With":[251],"share":[255],"domain,":[258],"it":[259],"performs":[260],"par":[262]},"counts_by_year":[],"updated_date":"2026-11-06T12:05:46.655746","created_date":"2026-11-04T00:00:00"}