{"id":"https://openalex.org/W7162294013","doi":"https://doi.org/10.48550/arxiv.2605.23067","title":"What Training Data Teaches RL Memory Agents: An Empirical Study of Curriculum Effects in Memory-Augmented QA","display_name":"What Training Data Teaches RL Memory Agents: An Empirical Study of Curriculum Effects in Memory-Augmented QA","publication_year":2026,"publication_date":"2026-05-21","ids":{"openalex":"https://openalex.org/W7162294013","doi":"https://doi.org/10.48550/arxiv.2605.23067"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.23067","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.23067","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":"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.23067","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136907891","display_name":"Xinjie He","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"He, Xinjie","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136985914","display_name":"Zhiyuan Lin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lin, Zhiyuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136910128","display_name":"Su Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Su","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102841028","display_name":"Jialun Wu","orcid":"https://orcid.org/0000-0002-9015-7487"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Jialun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136961264","display_name":"Qiyang Xie","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xie, Qiyang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136989535","display_name":"Weikai Zhou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhou, Weikai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5101193639","display_name":"Shuai Xiao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiao, Shuai","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/T10028","display_name":"Topic Modeling","score":0.18880000710487366,"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":0.18880000710487366,"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":0.14800000190734863,"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/T12808","display_name":"Ferroelectric and Negative Capacitance Devices","score":0.10040000081062317,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/curriculum","display_name":"Curriculum","score":0.6241000294685364},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.5774999856948853},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.5367000102996826},{"id":"https://openalex.org/keywords/empirical-research","display_name":"Empirical research","score":0.5220000147819519},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.4528999924659729},{"id":"https://openalex.org/keywords/aggregate","display_name":"Aggregate (composite)","score":0.4366999864578247},{"id":"https://openalex.org/keywords/experiential-learning","display_name":"Experiential learning","score":0.4278999865055084},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.3882000148296356},{"id":"https://openalex.org/keywords/lever","display_name":"Lever","score":0.387800008058548}],"concepts":[{"id":"https://openalex.org/C47177190","wikidata":"https://www.wikidata.org/wiki/Q207137","display_name":"Curriculum","level":2,"score":0.6241000294685364},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5900999903678894},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.5774999856948853},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.5367000102996826},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.5220000147819519},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.4528999924659729},{"id":"https://openalex.org/C4679612","wikidata":"https://www.wikidata.org/wiki/Q866298","display_name":"Aggregate (composite)","level":2,"score":0.4366999864578247},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4334000051021576},{"id":"https://openalex.org/C37228920","wikidata":"https://www.wikidata.org/wiki/Q1307600","display_name":"Experiential learning","level":2,"score":0.4278999865055084},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4146000146865845},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.3882000148296356},{"id":"https://openalex.org/C107524782","wikidata":"https://www.wikidata.org/wiki/Q40164","display_name":"Lever","level":2,"score":0.387800008058548},{"id":"https://openalex.org/C145420912","wikidata":"https://www.wikidata.org/wiki/Q853077","display_name":"Mathematics education","level":1,"score":0.3781999945640564},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.3659000098705292},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.36480000615119934},{"id":"https://openalex.org/C180747234","wikidata":"https://www.wikidata.org/wiki/Q23373","display_name":"Cognitive psychology","level":1,"score":0.34950000047683716},{"id":"https://openalex.org/C2778058735","wikidata":"https://www.wikidata.org/wiki/Q4692253","display_name":"Aggregate data","level":2,"score":0.3467000126838684},{"id":"https://openalex.org/C8642999","wikidata":"https://www.wikidata.org/wiki/Q4171168","display_name":"Hyperparameter","level":2,"score":0.3352000117301941},{"id":"https://openalex.org/C2777210771","wikidata":"https://www.wikidata.org/wiki/Q4927124","display_name":"Block (permutation group theory)","level":2,"score":0.31540000438690186},{"id":"https://openalex.org/C18762648","wikidata":"https://www.wikidata.org/wiki/Q42213","display_name":"Work (physics)","level":2,"score":0.3070000112056732},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.290800005197525},{"id":"https://openalex.org/C48372109","wikidata":"https://www.wikidata.org/wiki/Q3913","display_name":"Binary number","level":2,"score":0.28940001130104065},{"id":"https://openalex.org/C2777655017","wikidata":"https://www.wikidata.org/wiki/Q1501161","display_name":"Toolbox","level":2,"score":0.28630000352859497},{"id":"https://openalex.org/C67203356","wikidata":"https://www.wikidata.org/wiki/Q1321905","display_name":"Reinforcement","level":2,"score":0.27379998564720154},{"id":"https://openalex.org/C2778671685","wikidata":"https://www.wikidata.org/wiki/Q219239","display_name":"Recipe","level":2,"score":0.2734000086784363},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.259799987077713},{"id":"https://openalex.org/C138777275","wikidata":"https://www.wikidata.org/wiki/Q6884054","display_name":"Mixing (physics)","level":2,"score":0.257999986410141},{"id":"https://openalex.org/C136197465","wikidata":"https://www.wikidata.org/wiki/Q1729295","display_name":"Variety (cybernetics)","level":2,"score":0.25679999589920044},{"id":"https://openalex.org/C2781039887","wikidata":"https://www.wikidata.org/wiki/Q1391724","display_name":"Factor (programming language)","level":2,"score":0.25029999017715454}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.23067","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.23067","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.23067","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.23067","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":"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":[{"score":0.703141450881958,"display_name":"Quality Education","id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Reinforcement":[0],"learning":[1,180],"(RL)":[2],"has":[3],"emerged":[4],"as":[5,89],"a":[6,27,41,47,90,97,117,122,158],"viable":[7],"recipe":[8],"for":[9],"training":[10,36,64,172],"LLM":[11],"agents":[12],"to":[13,157,170],"reason":[14],"over":[15],"external":[16],"memory":[17,42,168],"banks":[18,169],"in":[19,198],"multi-session":[20],"dialogue.":[21],"Existing":[22],"work":[23],"trains":[24],"exclusively":[25],"on":[26,93,101,111,116,191],"single":[28],"benchmark,":[29],"leaving":[30],"open":[31],"how":[32],"the":[33,39,63,107,183],"composition":[34,87],"of":[35],"data":[37],"shapes":[38],"skills":[40],"agent":[43],"acquires.":[44],"We":[45,148],"present":[46],"controlled":[48],"empirical":[49],"study":[50],"that":[51,140],"holds":[52],"architecture,":[53],"RL":[54],"algorithm,":[55],"and":[56,60,75,82,174],"all":[57],"hyperparameters":[58],"fixed":[59],"varies":[61],"only":[62],"curriculum":[65,86,105,146],"across":[66],"three":[67],"conditions:":[68],"in-domain":[69],"(LoCoMo),":[70],"mixed-benchmark":[71],"(LoCoMo":[72],"+":[73],"LongMemEval),":[74],"out-of-domain":[76,119],"(LongMemEval":[77],"only).":[78],"Across":[79],"two":[80,151],"benchmarks":[81],"ten":[83],"question":[84],"types,":[85],"acts":[88],"fine-grained":[91],"lever":[92],"specialization":[94],"rather":[95],"than":[96],"uniform":[98],"scaling":[99],"factor":[100],"performance.":[102,132],"The":[103],"mixed":[104],"yields":[106],"strongest":[108],"overall":[109],"F1":[110],"both":[112],"evaluation":[113],"sets.":[114],"Training":[115],"narrow":[118],"set":[120],"transfers":[121],"targeted":[123],"skill":[124],"-":[125,128],"temporal":[126],"reasoning":[127],"despite":[129],"weak":[130],"aggregate":[131,137],"Per-type":[133],"differences":[134],"substantially":[135],"exceed":[136],"differences,":[138],"indicating":[139],"single-number":[141],"benchmark":[142],"comparisons":[143],"systematically":[144],"underreport":[145],"effects.":[147],"further":[149],"report":[150],"practical":[152],"lessons":[153],"from":[154,167],"adapting":[155],"GRPO":[156],"single-GPU":[159],"regime:":[160],"cross-benchmark":[161],"mixing":[162],"requires":[163],"filtering":[164],"format-specific":[165],"noise":[166],"preserve":[171],"signal,":[173],"binary":[175],"exact-match":[176],"reward":[177,196],"produces":[178],"no":[179],"signal":[181],"at":[182],"small":[184],"group":[185],"sizes":[186],"(G":[187],"=":[188],"4)":[189],"required":[190],"one":[192],"GPU,":[193],"motivating":[194],"continuous":[195],"functions":[197],"this":[199],"regime.":[200]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-26T00:00:00"}