{"id":"https://openalex.org/W7164015816","doi":"https://doi.org/10.48550/arxiv.2606.08347","title":"Tensorizing Engram: Sharing Latents Across N-Gram Embeddings is Beneficial in LLMs","display_name":"Tensorizing Engram: Sharing Latents Across N-Gram Embeddings is Beneficial in LLMs","publication_year":2026,"publication_date":"2026-06-06","ids":{"openalex":"https://openalex.org/W7164015816","doi":"https://doi.org/10.48550/arxiv.2606.08347"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.08347","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.08347","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.2606.08347","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5065141875","display_name":"Wuyang Zhou","orcid":"https://orcid.org/0000-0003-2229-2852"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhou, Wuyang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138236263","display_name":"Yuxuan Gu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gu, Yuxuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5094208275","display_name":"Giorgos Iacovides","orcid":"https://orcid.org/0009-0007-5733-8992"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Iacovides, Giorgos","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5070961417","display_name":"Yuning Qiu","orcid":"https://orcid.org/0000-0003-0268-0890"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qiu, Yuning","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138252398","display_name":"Qibin Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Qibin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5138245137","display_name":"Danilo Mandic","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mandic, Danilo","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":0.3919999897480011,"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":0.3919999897480011,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.3644999861717224,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.0478999987244606,"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/hash-function","display_name":"Hash function","score":0.6633999943733215},{"id":"https://openalex.org/keywords/transformer","display_name":"Transformer","score":0.5537999868392944},{"id":"https://openalex.org/keywords/encode","display_name":"ENCODE","score":0.459199994802475},{"id":"https://openalex.org/keywords/engram","display_name":"Engram","score":0.3849000036716461},{"id":"https://openalex.org/keywords/embedding","display_name":"Embedding","score":0.2791999876499176}],"concepts":[{"id":"https://openalex.org/C99138194","wikidata":"https://www.wikidata.org/wiki/Q183427","display_name":"Hash function","level":2,"score":0.6633999943733215},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6607999801635742},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.5537999868392944},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.47099998593330383},{"id":"https://openalex.org/C66746571","wikidata":"https://www.wikidata.org/wiki/Q1134833","display_name":"ENCODE","level":3,"score":0.459199994802475},{"id":"https://openalex.org/C28093023","wikidata":"https://www.wikidata.org/wiki/Q175168","display_name":"Engram","level":2,"score":0.3849000036716461},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.38029998540878296},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.289900004863739},{"id":"https://openalex.org/C41608201","wikidata":"https://www.wikidata.org/wiki/Q980509","display_name":"Embedding","level":2,"score":0.2791999876499176},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.2694000005722046},{"id":"https://openalex.org/C75291252","wikidata":"https://www.wikidata.org/wiki/Q1315756","display_name":"TRACE (psycholinguistics)","level":2,"score":0.25679999589920044},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2554999887943268}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.08347","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.08347","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.2606.08347","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.08347","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":{"Modern":[0],"language":[1],"models":[2],"represent":[3],"text":[4],"using":[5],"discrete":[6],"token-level":[7],"embeddings,":[8],"which":[9,48],"forces":[10],"recurring":[11],"multi-token":[12,34],"patterns":[13],"to":[14,27,98],"be":[15],"learned":[16],"implicitly":[17],"across":[18],"Transformer":[19],"layers.":[20],"Both":[21],"Over-tokenized":[22],"Transformers":[23],"and":[24,52],"Engram":[25,69],"attempt":[26],"address":[28,63],"this":[29],"limitation":[30],"by":[31],"explicitly":[32],"incorporating":[33],"(n-gram)":[35],"memories.":[36],"However,":[37],"they":[38],"rely":[39],"on":[40],"separate":[41],"hash":[42,50],"tables":[43],"for":[44],"each":[45],"n-gram":[46,78,104,116],"order,":[47],"introduces":[49],"collisions":[51],"prevents":[53],"nested":[54],"n-grams":[55],"from":[56],"sharing":[57],"the":[58,84,100],"underlying":[59],"latent":[60],"structures.":[61],"To":[62],"these":[64],"issues,":[65],"we":[66],"propose":[67],"Tensorized":[68],"(TN-gram),":[70],"a":[71],"compact":[72],"memory":[73],"module":[74],"that":[75,109],"represents":[76],"tensorized":[77],"embeddings":[79,101],"through":[80],"shared":[81,91],"factors":[82,93],"in":[83],"Canonical":[85],"Polyadic":[86],"(CP)":[87],"form.":[88],"TN-gram":[89,110],"learns":[90],"token-position":[92],"together":[94],"with":[95],"order-absorption":[96],"vectors":[97],"encode":[99],"of":[102],"different":[103],"order.":[105],"Comprehensive":[106],"experiments":[107],"demonstrate":[108],"matches":[111],"or":[112],"even":[113],"outperforms":[114],"Engram-style":[115],"modules":[117],"while":[118],"requiring":[119],"much":[120],"fewer":[121],"parameters.":[122]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-10T00:00:00"}