{"id":"https://openalex.org/W7214903209","doi":"https://doi.org/10.48550/arxiv.2609.36952","title":"ER-JEPA: Experience Replay Improves Joint-Embedding Predictive Learning in Language Models","display_name":"ER-JEPA: Experience Replay Improves Joint-Embedding Predictive Learning in Language Models","publication_year":2026,"publication_date":"2026-09-29","ids":{"openalex":"https://openalex.org/W7214903209","doi":"https://doi.org/10.48550/arxiv.2609.36952"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.36952","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.36952","pdf_url":"https://arxiv.org/pdf/2609.36952","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/2609.36952","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5153557395","display_name":"Jingnan Pu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I87445476","display_name":"Xi'an Jiaotong University","ror":"https://ror.org/017zhmm22","country_code":"CN","type":"education","lineage":["https://openalex.org/I87445476"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Pu, Jingnan","raw_affiliation_strings":["School of Automation Science and Technology , Xi'an Jiaotong University No. 28 , West Xianning Road , Xi'an , Shaanxi 710049 , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Automation Science and Technology , Xi'an Jiaotong University No. 28 , West Xianning Road , Xi'an , Shaanxi 710049 , China","institution_ids":["https://openalex.org/I87445476"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102616193","display_name":"Zi-En Fan","orcid":"https://orcid.org/0009-0008-7738-060X","observed_orcids":["https://orcid.org/0009-0008-7738-060X"]},"institutions":[{"id":"https://openalex.org/I87445476","display_name":"Xi'an Jiaotong University","ror":"https://ror.org/017zhmm22","country_code":"CN","type":"education","lineage":["https://openalex.org/I87445476"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Fan, Zi-En","raw_affiliation_strings":["School of Automation Science and Technology , Xi'an Jiaotong University No. 28 , West Xianning Road , Xi'an , Shaanxi 710049 , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Automation Science and Technology , Xi'an Jiaotong University No. 28 , West Xianning Road , Xi'an , Shaanxi 710049 , China","institution_ids":["https://openalex.org/I87445476"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5153654509","display_name":"Feng Lian","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I87445476","display_name":"Xi'an Jiaotong University","ror":"https://ror.org/017zhmm22","country_code":"CN","type":"education","lineage":["https://openalex.org/I87445476"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Lian, Feng","raw_affiliation_strings":["School of Automation Science and Technology , Xi'an Jiaotong University No. 28 , West Xianning Road , Xi'an , Shaanxi 710049 , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Automation Science and Technology , Xi'an Jiaotong University No. 28 , West Xianning Road , Xi'an , Shaanxi 710049 , China","institution_ids":["https://openalex.org/I87445476"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I87445476"],"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/T10028","display_name":"Topic Modeling","score":0.9253000020980835,"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":0.9253000020980835,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.027799999341368675,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.02669999934732914,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/experience-replay","display_name":"experience replay","score":0.722000002861023},{"id":"https://openalex.org/keywords/jepa","display_name":"JEPA","score":0.8999999761581421},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.8889999985694885},{"id":"https://openalex.org/keywords/representation-alignment","display_name":"representation alignment","score":0.4449999928474426},{"id":"https://openalex.org/keywords/episodic-memory","display_name":"episodic memory","score":0.8489999771118164},{"id":"https://openalex.org/keywords/token-prediction","display_name":"token prediction","score":0.6119999885559082},{"id":"https://openalex.org/keywords/knowledge-distillation","display_name":"knowledge distillation","score":0.41600000858306885},{"id":"https://openalex.org/keywords/language-model-fine-tuning","display_name":"language model fine-tuning","score":0.5649999976158142},{"id":"https://openalex.org/keywords/multi-view-alignment","display_name":"multi-view alignment","score":0.36500000953674316}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8259999752044678},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6394000053405762},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.5753999948501587},{"id":"https://openalex.org/C2983448237","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Language understanding","level":2,"score":0.5521000027656555},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.527400016784668},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.5180000066757202},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.5019000172615051},{"id":"https://openalex.org/C2777735758","wikidata":"https://www.wikidata.org/wiki/Q817765","display_name":"Path (computing)","level":2,"score":0.48590001463890076},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.4722000062465668},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.4706999957561493},{"id":"https://openalex.org/C61797465","wikidata":"https://www.wikidata.org/wiki/Q1188986","display_name":"Term (time)","level":2,"score":0.3723999857902527},{"id":"https://openalex.org/C161301231","wikidata":"https://www.wikidata.org/wiki/Q3478658","display_name":"Knowledge representation and reasoning","level":2,"score":0.35510000586509705},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.3515999913215637},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.3456000089645386},{"id":"https://openalex.org/C67186912","wikidata":"https://www.wikidata.org/wiki/Q367664","display_name":"Data modeling","level":2,"score":0.3379000127315521},{"id":"https://openalex.org/C74672266","wikidata":"https://www.wikidata.org/wiki/Q815859","display_name":"Language acquisition","level":2,"score":0.30570000410079956},{"id":"https://openalex.org/C2780440489","wikidata":"https://www.wikidata.org/wiki/Q5227278","display_name":"Data-driven","level":2,"score":0.2985000014305115},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.28450000286102295},{"id":"https://openalex.org/C59404180","wikidata":"https://www.wikidata.org/wiki/Q17013334","display_name":"Feature learning","level":2,"score":0.2791000008583069},{"id":"https://openalex.org/C45804977","wikidata":"https://www.wikidata.org/wiki/Q7239673","display_name":"Predictive modelling","level":2,"score":0.27230000495910645},{"id":"https://openalex.org/C116409475","wikidata":"https://www.wikidata.org/wiki/Q1385056","display_name":"External Data Representation","level":2,"score":0.25699999928474426}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.36952","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.36952","pdf_url":"https://arxiv.org/pdf/2609.36952","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.2609.36952","doi":"https://doi.org/10.48550/arxiv.2609.36952","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.36952","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:2609.36952","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.36952","pdf_url":"https://arxiv.org/pdf/2609.36952","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/W7214903209.pdf","grobid_xml":"https://content.openalex.org/works/W7214903209.grobid-xml"},"referenced_works_count":13,"referenced_works":["https://openalex.org/W2947497897","https://openalex.org/W3198002980","https://openalex.org/W3210904070","https://openalex.org/W4303443398","https://openalex.org/W4392427358","https://openalex.org/W4392677872","https://openalex.org/W4394861491","https://openalex.org/W4401306886","https://openalex.org/W4402672081","https://openalex.org/W4405903187","https://openalex.org/W4410371768","https://openalex.org/W4414888753","https://openalex.org/W4417261359"],"related_works":["https://openalex.org/W7125153218","https://openalex.org/W7163645536","https://openalex.org/W4415798561","https://openalex.org/W7139076825","https://openalex.org/W7161559799","https://openalex.org/W7152933663","https://openalex.org/W4412887988","https://openalex.org/W4416512539","https://openalex.org/W7131904352","https://openalex.org/W7171529859"],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2],"(LLMs)":[3],"excel":[4],"at":[5],"token-level":[6],"generation":[7],"but":[8],"may":[9],"learn":[10],"undesirable":[11],"abstract":[12],"semantics":[13],"and":[14,73,89,99,109],"lack":[15],"comprehensive":[16],"perception.":[17],"LLM-JEPA":[18],"mitigates":[19],"this":[20],"by":[21],"aligning":[22],"different":[23],"views":[24],"of":[25],"the":[26,86],"same":[27],"underlying":[28],"knowledge":[29],"via":[30],"a":[31,66],"joint-embedding":[32],"predictive":[33],"architecture":[34],"(JEPA).":[35],"However,":[36],"strong":[37],"alignment":[38],"does":[39],"not":[40],"necessarily":[41],"lead":[42],"to":[43,59,77],"accurate,":[44],"stable":[45],"predictions.":[46],"To":[47],"address":[48],"this,":[49],"we":[50],"propose":[51],"ER-JEPA,":[52],"which":[53],"adds":[54],"an":[55],"episodic":[56],"replay":[57],"path":[58],"LLM-JEPA.":[60,116],"ER-JEPA":[61,113],"stores":[62,72],"training":[63,91],"pairs":[64],"in":[65],"memory.":[67],"At":[68],"each":[69],"step,":[70],"it":[71],"retrieves":[74],"relevant":[75],"data":[76],"provide":[78],"additional":[79,94],"supervision.":[80],"This":[81],"enables":[82],"learning":[83],"from":[84],"both":[85],"current":[87],"batch":[88],"stored":[90],"pairs,":[92],"providing":[93],"supervision":[95],"for":[96],"token":[97],"prediction":[98],"representation":[100],"alignment.":[101],"Experiments":[102],"across":[103],"multiple":[104],"datasets":[105],"(NL-RX,":[106],"GSM8K,":[107],"Spider,":[108],"NQ-Open)":[110],"demonstrate":[111],"that":[112],"consistently":[114],"outperforms":[115]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-10-01T00:00:00"}