{"id":"https://openalex.org/W7203924436","doi":"https://doi.org/10.48550/arxiv.2608.18827","title":"MLREF: Efficient Module Reuse for Reward Design in Reinforcement Learning via Large Language Models","display_name":"MLREF: Efficient Module Reuse for Reward Design in Reinforcement Learning via Large Language Models","publication_year":2026,"publication_date":"2026-08-19","ids":{"openalex":"https://openalex.org/W7203924436","doi":"https://doi.org/10.48550/arxiv.2608.18827"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.18827","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.18827","pdf_url":"https://arxiv.org/pdf/2608.18827","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","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.18827","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5082084042","display_name":"\u5218\u6210\u6797","orcid":"https://orcid.org/0000-0001-9032-2991","observed_orcids":["https://orcid.org/0000-0001-9032-2991"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Liu, Chenglin","raw_affiliation_strings":["Institute for Interdisciplinary Information Sciences , Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute for Interdisciplinary Information Sciences , Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126627812","display_name":"Xun Wang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wang, Xun","raw_affiliation_strings":["Institute for Interdisciplinary Information Sciences , Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute for Interdisciplinary Information Sciences , Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130235069","display_name":"Ruishuo Chen","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chen, Ruishuo","raw_affiliation_strings":["Institute for Interdisciplinary Information Sciences , Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute for Interdisciplinary Information Sciences , Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5014772321","display_name":"Zhuoran Li","orcid":"https://orcid.org/0000-0002-8776-9562","observed_orcids":["https://orcid.org/0000-0002-8776-9562"]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Li, Zhuoran","raw_affiliation_strings":["Institute for Interdisciplinary Information Sciences , Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute for Interdisciplinary Information Sciences , Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5147700071","display_name":"Longbo Huang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I99065089","display_name":"Tsinghua University","ror":"https://ror.org/03cve4549","country_code":"CN","type":"education","lineage":["https://openalex.org/I99065089"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Huang, Longbo","raw_affiliation_strings":["Institute for Interdisciplinary Information Sciences , Tsinghua University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute for Interdisciplinary Information Sciences , Tsinghua University","institution_ids":["https://openalex.org/I99065089"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I99065089"],"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.9994999766349792,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.9994999766349792,"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/T11975","display_name":"Evolutionary Algorithms and Applications","score":0.00019999999494757503,"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/T10906","display_name":"AI-based Problem Solving and Planning","score":0.0,"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/reward-function-design","display_name":"reward function design","score":0.8489999771118164},{"id":"https://openalex.org/keywords/module-reuse","display_name":"module reuse","score":0.8980000019073486},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"reinforcement learning","score":0.6290000081062317}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7623000144958496},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.754800021648407},{"id":"https://openalex.org/C206588197","wikidata":"https://www.wikidata.org/wiki/Q846574","display_name":"Reuse","level":2,"score":0.7378000020980835},{"id":"https://openalex.org/C2780513914","wikidata":"https://www.wikidata.org/wiki/Q18210350","display_name":"Bottleneck","level":2,"score":0.6729000210762024},{"id":"https://openalex.org/C197129107","wikidata":"https://www.wikidata.org/wiki/Q1921621","display_name":"Merge (version control)","level":2,"score":0.6632999777793884},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.5867000222206116},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4194999933242798},{"id":"https://openalex.org/C14036430","wikidata":"https://www.wikidata.org/wiki/Q3736076","display_name":"Function (biology)","level":2,"score":0.3353999853134155},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.32510000467300415},{"id":"https://openalex.org/C2777904410","wikidata":"https://www.wikidata.org/wiki/Q7397","display_name":"Software","level":2,"score":0.29910001158714294},{"id":"https://openalex.org/C2778583558","wikidata":"https://www.wikidata.org/wiki/Q771245","display_name":"Code reuse","level":3,"score":0.26649999618530273},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.25609999895095825},{"id":"https://openalex.org/C67203356","wikidata":"https://www.wikidata.org/wiki/Q1321905","display_name":"Reinforcement","level":2,"score":0.2551000118255615},{"id":"https://openalex.org/C168167062","wikidata":"https://www.wikidata.org/wiki/Q1117970","display_name":"Component (thermodynamics)","level":2,"score":0.250900000333786}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.18827","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.18827","pdf_url":"https://arxiv.org/pdf/2608.18827","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.2608.18827","doi":"https://doi.org/10.48550/arxiv.2608.18827","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.18827","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:2608.18827","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.18827","pdf_url":"https://arxiv.org/pdf/2608.18827","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":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7203924436.pdf","grobid_xml":"https://content.openalex.org/works/W7203924436.grobid-xml"},"referenced_works_count":9,"referenced_works":["https://openalex.org/W2807192342","https://openalex.org/W2980897231","https://openalex.org/W3119264617","https://openalex.org/W3177813494","https://openalex.org/W3193576605","https://openalex.org/W4322825501","https://openalex.org/W4380993732","https://openalex.org/W7106169679","https://openalex.org/W7165385035"],"related_works":["https://openalex.org/W4403996116","https://openalex.org/W7169839044","https://openalex.org/W7172494134","https://openalex.org/W4399401029","https://openalex.org/W7106695871","https://openalex.org/W7154617860","https://openalex.org/W4414742936","https://openalex.org/W4414792178","https://openalex.org/W6967162250","https://openalex.org/W7133342563"],"abstract_inverted_index":{"Reward":[0,56],"function":[1],"design":[2],"remains":[3],"a":[4,66,69,130],"bottleneck":[5],"in":[6,40,157,161],"reinforcement":[7],"learning.":[8],"While":[9],"large":[10],"language":[11],"models":[12],"(LLMs)":[13],"have":[14],"enabled":[15],"automated":[16],"reward":[17,24,74,103,143],"generation,":[18],"existing":[19],"methods":[20],"generate":[21],"and":[22,35,98,129,140,159],"revise":[23],"functions":[25,104],"as":[26,81,107],"monolithic":[27],"programs,":[28],"making":[29],"it":[30],"difficult":[31],"to":[32,44],"reliably":[33],"preserve":[34],"reuse":[36],"effective":[37],"components":[38],"discovered":[39],"earlier":[41],"iterations,":[42],"leading":[43],"unstable":[45],"performance":[46],"across":[47,89],"iterations.":[48],"To":[49,116],"address":[50],"this,":[51],"we":[52],"propose":[53],"Module":[54],"Level":[55],"Evolution":[57],"Framework":[58],"(MLREF).":[59],"At":[60],"the":[61,78,82,86,138],"core":[62],"of":[63,72,110,142],"MLREF":[64,76,120,151],"is":[65],"module":[67,79],"pool,":[68],"persistent":[70],"repository":[71],"reusable":[73],"components.":[75],"treats":[77],"pool":[80,87],"primary":[83],"optimization":[84,166],"object:":[85],"evolves":[88],"iterations":[90],"by":[91,155],"accumulating":[92],"successful":[93],"modules,":[94],"refining":[95],"underperforming":[96],"ones,":[97],"reusing":[99],"proven":[100],"components;":[101],"while":[102],"are":[105],"constructed":[106],"linear":[108],"combinations":[109],"modules":[111],"drawn":[112],"from":[113],"this":[114,118],"pool.":[115],"drive":[117],"evolution,":[119],"integrates":[121],"three":[122],"mechanisms:":[123],"reflection-based":[124],"refinement,":[125],"hybrid":[126],"credit":[127],"assignment,":[128],"merge":[131],"strategy":[132],"with":[133,163],"rollback,":[134],"which":[135],"together":[136],"improve":[137],"effectiveness":[139],"robustness":[141],"optimization.":[144],"Experiments":[145],"on":[146],"17":[147],"tasks":[148],"show":[149],"that":[150],"outperforms":[152],"strong":[153],"baselines":[154],"25.2%":[156],"locomotion":[158],"6.6%":[160],"manipulation,":[162],"more":[164],"stable":[165],"dynamics.":[167]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-08-22T00:00:00"}