{"id":"https://openalex.org/W7160920911","doi":"https://doi.org/10.48550/arxiv.2605.08809","title":"SimReg: Achieving Higher Performance in the Pretraining via Embedding Similarity Regularization","display_name":"SimReg: Achieving Higher Performance in the Pretraining via Embedding Similarity Regularization","publication_year":2026,"publication_date":"2026-05-09","ids":{"openalex":"https://openalex.org/W7160920911","doi":"https://doi.org/10.48550/arxiv.2605.08809"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.08809","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.08809","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.2605.08809","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5135979744","display_name":"Yan Sun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Yan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135980295","display_name":"Guoxia Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Guoxia","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5035973185","display_name":"Jinle Zeng","orcid":"https://orcid.org/0000-0002-8742-3553"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zeng, Jinle","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135928701","display_name":"JiaBin Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, JiaBin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135945259","display_name":"Shuai Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Shuai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135979288","display_name":"Li Shen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shen, Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135996034","display_name":"Dacheng Tao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tao, Dacheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5084155236","display_name":"Dianhai Yu","orcid":"https://orcid.org/0000-0002-0163-2603"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yu, DianHai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5135992874","display_name":"Haifeng Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Haifeng","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":0.3977000117301941,"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":0.3977000117301941,"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":0.11710000038146973,"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/T13702","display_name":"Machine Learning in Healthcare","score":0.0706000030040741,"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/regularization","display_name":"Regularization (linguistics)","score":0.6809999942779541},{"id":"https://openalex.org/keywords/embedding","display_name":"Embedding","score":0.673799991607666},{"id":"https://openalex.org/keywords/hyperparameter","display_name":"Hyperparameter","score":0.6241000294685364},{"id":"https://openalex.org/keywords/similarity","display_name":"Similarity (geometry)","score":0.5253999829292297},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.41940000653266907},{"id":"https://openalex.org/keywords/pattern-recognition","display_name":"Pattern recognition (psychology)","score":0.4090000092983246}],"concepts":[{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.6809999942779541},{"id":"https://openalex.org/C41608201","wikidata":"https://www.wikidata.org/wiki/Q980509","display_name":"Embedding","level":2,"score":0.673799991607666},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6320000290870667},{"id":"https://openalex.org/C8642999","wikidata":"https://www.wikidata.org/wiki/Q4171168","display_name":"Hyperparameter","level":2,"score":0.6241000294685364},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5866000056266785},{"id":"https://openalex.org/C103278499","wikidata":"https://www.wikidata.org/wiki/Q254465","display_name":"Similarity (geometry)","level":3,"score":0.5253999829292297},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.41940000653266907},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.41190001368522644},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.4090000092983246},{"id":"https://openalex.org/C196083921","wikidata":"https://www.wikidata.org/wiki/Q7915758","display_name":"Variance (accounting)","level":2,"score":0.3684000074863434},{"id":"https://openalex.org/C59404180","wikidata":"https://www.wikidata.org/wiki/Q17013334","display_name":"Feature learning","level":2,"score":0.3296000063419342},{"id":"https://openalex.org/C2777303404","wikidata":"https://www.wikidata.org/wiki/Q759757","display_name":"Convergence (economics)","level":2,"score":0.32820001244544983},{"id":"https://openalex.org/C94124525","wikidata":"https://www.wikidata.org/wiki/Q912550","display_name":"Categorization","level":2,"score":0.32409998774528503},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.3052000105381012},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.2816999852657318},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.27000001072883606}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.08809","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.08809","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.2605.08809","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.08809","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":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Pretraining":[0],"large":[1],"language":[2],"models":[3,22],"(LLMs)":[4],"with":[5,77],"next-token":[6],"prediction":[7],"has":[8,41],"led":[9],"to":[10,85],"remarkable":[11],"advances,":[12],"yet":[13],"the":[14,33,65,78],"context-dependent":[15],"nature":[16],"of":[17,35],"token":[18,75],"embeddings":[19],"in":[20,24,44,54],"such":[21],"results":[23],"high":[25],"intra-class":[26],"variance":[27],"and":[28,47,52,120,134,149,157],"inter-class":[29],"similarity,":[30],"thus":[31],"hindering":[32],"efficiency":[34],"representation":[36],"learning.":[37],"While":[38],"similarity-based":[39],"regularization":[40,70],"demonstrated":[42],"benefit":[43],"supervised":[45],"fine-tuning":[46],"classification":[48],"tasks,":[49],"its":[50],"application":[51],"efficacy":[53],"large-scale":[55],"LLM":[56],"pretraining":[57],"remains":[58],"underexplored.":[59],"In":[60],"this":[61,103],"work,":[62],"we":[63],"propose":[64],"SimReg,":[66],"an":[67],"embedding":[68],"similarity":[69],"loss":[71,158],"that":[72,102,125],"explicitly":[73],"encourages":[74],"representations":[76],"same":[79],"ground-truth":[80],"label":[81],"within":[82],"each":[83],"sequence":[84],"be":[86],"more":[87,113],"similar,":[88],"while":[89],"enforcing":[90],"separation":[91],"from":[92],"different-label":[93],"tokens":[94],"via":[95],"a":[96],"contrastive":[97],"loss.":[98],"Our":[99],"analysis":[100],"reveals":[101],"mechanism":[104],"introduces":[105],"gains":[106],"by":[107,131,140],"enlarging":[108],"multi-classification":[109],"margins,":[110],"thereby":[111],"enabling":[112],"efficient":[114],"classification.":[115],"Extensive":[116],"experiments":[117],"across":[118,143],"dense":[119],"Mixture-of-Experts":[121],"(MoE)":[122],"architectures":[123],"demonstrate":[124],"SimReg":[126],"consistently":[127],"accelerates":[128],"training":[129],"convergence":[130],"over":[132,141],"30%":[133],"improves":[135],"average":[136],"zero-shot":[137],"downstream":[138],"performance":[139],"1%":[142],"standard":[144],"benchmarks.":[145],"Further":[146],"ablation":[147],"studies":[148],"analyses":[150],"offer":[151],"practical":[152],"insights":[153],"into":[154],"hyperparameter":[155],"tuning":[156],"effectiveness.":[159]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-13T00:00:00"}