{"id":"https://openalex.org/W7165621090","doi":"https://doi.org/10.48550/arxiv.2606.21089","title":"Repeated post-training is not Self-improving: Diagnosing Scientific Amnesia in Continual DPO Pipelines","display_name":"Repeated post-training is not Self-improving: Diagnosing Scientific Amnesia in Continual DPO Pipelines","publication_year":2026,"publication_date":"2026-06-17","ids":{"openalex":"https://openalex.org/W7165621090","doi":"https://doi.org/10.48550/arxiv.2606.21089"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.21089","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.21089","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2606.21089","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101648582","display_name":"Jianzhe Lin","orcid":"https://orcid.org/0000-0002-1559-1048"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lin, Jianzhe","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5029787563","display_name":"Fang Wang","orcid":"https://orcid.org/0000-0002-3327-4177"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Fei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139142713","display_name":"Xiaolin Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Xiaolin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120024901","display_name":"Rajeshkumar Golani","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Golani, Rajeshkumar","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5093932120","display_name":"Jubin Chheda","orcid":"https://orcid.org/0009-0005-0311-7890"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chheda, Jubin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"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/T10906","display_name":"AI-based Problem Solving and Planning","score":0.46129998564720154,"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/T10906","display_name":"AI-based Problem Solving and Planning","score":0.46129998564720154,"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/T10260","display_name":"Software Engineering Research","score":0.04349999874830246,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/T11596","display_name":"Constraint Satisfaction and Optimization","score":0.039400000125169754,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"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/intuition","display_name":"Intuition","score":0.46239998936653137},{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.41769999265670776},{"id":"https://openalex.org/keywords/suite","display_name":"Suite","score":0.38749998807907104},{"id":"https://openalex.org/keywords/semantic-reasoner","display_name":"Semantic reasoner","score":0.3813000023365021},{"id":"https://openalex.org/keywords/pipeline-transport","display_name":"Pipeline transport","score":0.3319999873638153},{"id":"https://openalex.org/keywords/probabilistic-logic","display_name":"Probabilistic logic","score":0.3246999979019165},{"id":"https://openalex.org/keywords/amnesia","display_name":"Amnesia","score":0.3163999915122986},{"id":"https://openalex.org/keywords/causal-reasoning","display_name":"Causal reasoning","score":0.304500013589859}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4912000000476837},{"id":"https://openalex.org/C132010649","wikidata":"https://www.wikidata.org/wiki/Q189222","display_name":"Intuition","level":2,"score":0.46239998936653137},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.41769999265670776},{"id":"https://openalex.org/C79581498","wikidata":"https://www.wikidata.org/wiki/Q1367530","display_name":"Suite","level":2,"score":0.38749998807907104},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.3824999928474426},{"id":"https://openalex.org/C9616225","wikidata":"https://www.wikidata.org/wiki/Q3929429","display_name":"Semantic reasoner","level":2,"score":0.3813000023365021},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.37310001254081726},{"id":"https://openalex.org/C175309249","wikidata":"https://www.wikidata.org/wiki/Q725864","display_name":"Pipeline transport","level":2,"score":0.3319999873638153},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.3276999890804291},{"id":"https://openalex.org/C49937458","wikidata":"https://www.wikidata.org/wiki/Q2599292","display_name":"Probabilistic logic","level":2,"score":0.3246999979019165},{"id":"https://openalex.org/C2778615406","wikidata":"https://www.wikidata.org/wiki/Q11072","display_name":"Amnesia","level":2,"score":0.3163999915122986},{"id":"https://openalex.org/C115086926","wikidata":"https://www.wikidata.org/wiki/Q17004651","display_name":"Causal reasoning","level":3,"score":0.304500013589859},{"id":"https://openalex.org/C4554734","wikidata":"https://www.wikidata.org/wiki/Q593744","display_name":"Knowledge base","level":2,"score":0.2957000136375427},{"id":"https://openalex.org/C46686674","wikidata":"https://www.wikidata.org/wiki/Q466303","display_name":"Boosting (machine learning)","level":2,"score":0.2946000099182129},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.26840001344680786},{"id":"https://openalex.org/C184898388","wikidata":"https://www.wikidata.org/wiki/Q1435712","display_name":"Pairwise comparison","level":2,"score":0.26809999346733093},{"id":"https://openalex.org/C77595967","wikidata":"https://www.wikidata.org/wiki/Q3151013","display_name":"Forensic engineering","level":1,"score":0.26669999957084656},{"id":"https://openalex.org/C66283442","wikidata":"https://www.wikidata.org/wiki/Q1389268","display_name":"Failure mode and effects analysis","level":2,"score":0.26440000534057617},{"id":"https://openalex.org/C68387754","wikidata":"https://www.wikidata.org/wiki/Q7271585","display_name":"Schedule","level":2,"score":0.2590999901294708},{"id":"https://openalex.org/C2779119184","wikidata":"https://www.wikidata.org/wiki/Q294350","display_name":"ALARM","level":2,"score":0.257999986410141},{"id":"https://openalex.org/C2776360696","wikidata":"https://www.wikidata.org/wiki/Q1938414","display_name":"Taxpayer","level":2,"score":0.25690001249313354},{"id":"https://openalex.org/C66882249","wikidata":"https://www.wikidata.org/wiki/Q169336","display_name":"Homogeneous","level":2,"score":0.2551000118255615},{"id":"https://openalex.org/C2777019708","wikidata":"https://www.wikidata.org/wiki/Q1412340","display_name":"Irrationality","level":3,"score":0.2540999948978424},{"id":"https://openalex.org/C188198153","wikidata":"https://www.wikidata.org/wiki/Q1613840","display_name":"Limiting","level":2,"score":0.2531000077724457},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.2522999942302704}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.21089","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.21089","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.21089","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.21089","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Industrial":[0],"LLM":[1],"teams":[2],"often":[3],"ship":[4],"behavior":[5],"updates":[6],"by":[7],"repeatedly":[8],"DPO-training":[9],"a":[10,32,68,75,81,93,100,119,126,160,173,198,210],"base":[11],"model":[12],"on":[13,220],"sequences":[14],"of":[15,104,134],"related":[16],"preference-data":[17],"campaigns.":[18],"The":[19,192],"dominant":[20],"failure":[21,57],"mode":[22,58],"in":[23,138,159,172,209],"this":[24,56,157],"regime":[25],"is":[26,164,189,194,207],"not":[27,197],"always":[28],"classical":[29],"catastrophic":[30],"forgetting:":[31],"pipeline":[33,83],"may":[34],"preserve":[35],"previously":[36],"learned":[37],"behaviors":[38],"while":[39],"still":[40],"failing":[41],"to":[42,49],"accumulate":[43],"reusable":[44],"methodological":[45],"knowledge":[46],"about":[47,216],"how":[48],"train":[50],"the":[51,145,165,180,203],"next":[52],"campaign.":[53],"We":[54,72],"call":[55],"scientific":[59,205],"amnesia.":[60],"This":[61],"paper":[62],"turns":[63],"that":[64,84,169,200],"practitioner":[65],"intuition":[66],"into":[67],"measurable":[69],"industrial":[70],"problem.":[71],"contribute:":[73],"(i)":[74],"diagnostic":[76,102],"suite":[77],"for":[78],"amnesia,":[79],"(ii)":[80],"Program-based":[82],"chains":[85],"FSDP-sharded":[86],"DPO":[87],"checkpoints":[88],"across":[89],"Qwen2.5-7B-Instruct":[90],"runs,":[91],"(iii)":[92],"30-campaign":[94],"HumanEval":[95],"subdomain":[96],"benchmark,":[97],"and":[98,117,122,184,214,225],"(iv)":[99],"comparative":[101],"study":[103],"five":[105],"strategy":[106],"proposers:":[107],"random":[108],"memory,":[109,113],"rule-based":[110,148],"scheduling,":[111],"retrieval-only":[112,178],"warm-start":[114],"Bayesian":[115],"optimization,":[116],"MSCL,":[118],"meta-scientific":[120],"memory":[121],"reasoner":[123],"candidate.":[124],"Across":[125],"single-seed":[127],"5-condition":[128],"*":[129],"3-step":[130],"real-LM":[131],"chain,":[132,162],"4":[133],"5":[135],"candidates":[136],"degrade":[137],"step-level":[139],"peak":[140],"pass@1,":[141],"including":[142],"MSCL;":[143],"only":[144,166],"deliberately":[146],"conservative":[147],"schedule":[149],"improves.":[150],"Follow-up":[151],"pilots":[152],"qualify":[153],"rather":[154],"than":[155],"overturn":[156],"finding:":[158],"heterogeneous":[161],"MSCL":[163,201],"completed":[167],"candidate":[168,187],"improves,":[170],"whereas":[171],"small":[174],"multi-seed":[175],"homogeneous":[176],"sweep,":[177],"has":[179],"best":[181],"mean":[182],"Delta":[183],"no":[185],"pairwise":[186],"gap":[188],"statistically":[190],"distinguishable.":[191],"contribution":[193],"therefore":[195],"diagnostic,":[196],"claim":[199],"solves":[202],"problem:":[204],"amnesia":[206],"observable":[208],"production-like":[211],"continual-DPO":[212],"pipeline,":[213],"conclusions":[215],"interventions":[217],"depend":[218],"sharply":[219],"chain":[221],"regime,":[222],"evaluator":[223],"design,":[224],"seed":[226],"coverage.":[227]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-24T00:00:00"}