{"id":"https://openalex.org/W7154758010","doi":"https://doi.org/10.18653/2026.emnlp-main.19","title":"Compressing Sequences in the Latent Embedding Space: K-Token Merging for Large Language Models","display_name":"Compressing Sequences in the Latent Embedding Space: K-Token Merging for Large Language Models","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7154758010","doi":"https://doi.org/10.18653/2026.emnlp-main.19"},"language":"en","primary_location":{"id":"doi:10.18653/2026.emnlp-main.19","doi":"https://doi.org/10.18653/2026.emnlp-main.19","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.emnlp-main.19","pdf_url":"https://aclanthology.org/2026.emnlp-main.19.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"bronze","oa_url":"https://aclanthology.org/2026.emnlp-main.19.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133923303","display_name":"Zihao Xu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4210096112","display_name":"Rutgers Sexual and Reproductive Health and Rights","ror":"https://ror.org/00rcvgx40","country_code":"NL","type":"other","lineage":["https://openalex.org/I4210096112"]}],"countries":["NL"],"is_corresponding":false,"raw_author_name":"Zihao Xu","raw_affiliation_strings":["Rutgers University , \u2660 AWS AI Labs ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Rutgers University , \u2660 AWS AI Labs ,","institution_ids":["https://openalex.org/I4210096112"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5078256928","display_name":"John B. Harvill","orcid":"https://orcid.org/0000-0003-3633-6756","observed_orcids":["https://orcid.org/0000-0003-3633-6756"]},"institutions":[{"id":"https://openalex.org/I4210096112","display_name":"Rutgers Sexual and Reproductive Health and Rights","ror":"https://ror.org/00rcvgx40","country_code":"NL","type":"other","lineage":["https://openalex.org/I4210096112"]}],"countries":["NL"],"is_corresponding":false,"raw_author_name":"John Harvill","raw_affiliation_strings":["Rutgers University , \u2660 AWS AI Labs ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Rutgers University , \u2660 AWS AI Labs ,","institution_ids":["https://openalex.org/I4210096112"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5017213607","display_name":"Ziwei Fan","orcid":"https://orcid.org/0000-0001-5445-2203","observed_orcids":["https://orcid.org/0000-0001-5445-2203"]},"institutions":[{"id":"https://openalex.org/I4210096112","display_name":"Rutgers Sexual and Reproductive Health and Rights","ror":"https://ror.org/00rcvgx40","country_code":"NL","type":"other","lineage":["https://openalex.org/I4210096112"]}],"countries":["NL"],"is_corresponding":false,"raw_author_name":"Ziwei Fan","raw_affiliation_strings":["Rutgers University , \u2660 AWS AI Labs ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Rutgers University , \u2660 AWS AI Labs ,","institution_ids":["https://openalex.org/I4210096112"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126252841","display_name":"Yizhou Sun","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I157725225","display_name":"University of Illinois Urbana-Champaign","ror":"https://ror.org/047426m28","country_code":"US","type":"education","lineage":["https://openalex.org/I157725225"]},{"id":"https://openalex.org/I161318765","display_name":"University of California, Los Angeles","ror":"https://ror.org/046rm7j60","country_code":"US","type":"education","lineage":["https://openalex.org/I161318765"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yizhou Sun","raw_affiliation_strings":["University of California Los Angeles , \u25b3 University of Illinois Urbana-Champaign"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"University of California Los Angeles , \u25b3 University of Illinois Urbana-Champaign","institution_ids":["https://openalex.org/I157725225","https://openalex.org/I161318765"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133897863","display_name":"Hao Ding","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4210096112","display_name":"Rutgers Sexual and Reproductive Health and Rights","ror":"https://ror.org/00rcvgx40","country_code":"NL","type":"other","lineage":["https://openalex.org/I4210096112"]}],"countries":["NL"],"is_corresponding":false,"raw_author_name":"Hao Ding","raw_affiliation_strings":["Rutgers University , \u2660 AWS AI Labs ,"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Rutgers University , \u2660 AWS AI Labs ,","institution_ids":["https://openalex.org/I4210096112"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5133828998","display_name":"Hao Wang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hao Wang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.19584424,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"419","last_page":"429"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9550999999046326,"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":0.9550999999046326,"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":0.03229999914765358,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.004699999932199717,"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/token-compression","display_name":"token compression","score":0.652999997138977},{"id":"https://openalex.org/keywords/prompt-compression","display_name":"prompt compression","score":0.7680000066757202},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.746999979019165},{"id":"https://openalex.org/keywords/lora","display_name":"LoRa","score":0.6449999809265137},{"id":"https://openalex.org/keywords/structural-reasoning","display_name":"structural reasoning","score":0.6010000109672546},{"id":"https://openalex.org/keywords/sentiment-classification","display_name":"sentiment classification","score":0.9670000076293945},{"id":"https://openalex.org/keywords/source-code-modification","display_name":"code editing","score":0.9169999957084656}],"concepts":[{"id":"https://openalex.org/C41608201","wikidata":"https://www.wikidata.org/wiki/Q980509","display_name":"Embedding","level":2,"score":0.7685999870300293},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7447999715805054},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.7143999934196472},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.572700023651123},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.5393999814987183},{"id":"https://openalex.org/C2777210771","wikidata":"https://www.wikidata.org/wiki/Q4927124","display_name":"Block (permutation group theory)","level":2,"score":0.517799973487854},{"id":"https://openalex.org/C180016635","wikidata":"https://www.wikidata.org/wiki/Q2712821","display_name":"Compression (physics)","level":2,"score":0.5002999901771545},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.4936000108718872},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.4690999984741211},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.45750001072883606},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.4381999969482422},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4090999960899353},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.3977000117301941},{"id":"https://openalex.org/C137635306","wikidata":"https://www.wikidata.org/wiki/Q182667","display_name":"Pareto principle","level":2,"score":0.35510000586509705},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3467000126838684},{"id":"https://openalex.org/C2778572836","wikidata":"https://www.wikidata.org/wiki/Q380933","display_name":"Space (punctuation)","level":2,"score":0.34310001134872437},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.2849000096321106},{"id":"https://openalex.org/C2776848632","wikidata":"https://www.wikidata.org/wiki/Q853463","display_name":"Clipping (morphology)","level":2,"score":0.26460000872612},{"id":"https://openalex.org/C81081738","wikidata":"https://www.wikidata.org/wiki/Q55542","display_name":"Lossless compression","level":3,"score":0.25999999046325684},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.2572999894618988},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.25060001015663147}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/2026.emnlp-main.19","doi":"https://doi.org/10.18653/2026.emnlp-main.19","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.emnlp-main.19","pdf_url":"https://aclanthology.org/2026.emnlp-main.19.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2604.15153","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2604.15153","pdf_url":"https://arxiv.org/pdf/2604.15153","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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2604.15153","doi":"https://doi.org/10.48550/arxiv.2604.15153","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.15153","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":"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"}],"best_oa_location":{"id":"doi:10.18653/2026.emnlp-main.19","doi":"https://doi.org/10.18653/2026.emnlp-main.19","is_oa":true,"landing_page_url":"https://doi.org/10.18653/2026.emnlp-main.19","pdf_url":"https://aclanthology.org/2026.emnlp-main.19.pdf","source":{"id":"https://openalex.org/S4363608991","display_name":"Conference on Empirical Methods in Natural Language Processing (EMNLP)","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7154758010.pdf","grobid_xml":"https://content.openalex.org/works/W7154758010.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4414889823","https://openalex.org/W7111251111","https://openalex.org/W4414374067","https://openalex.org/W4407124183","https://openalex.org/W4389519226","https://openalex.org/W4402954149","https://openalex.org/W4416354534","https://openalex.org/W4402978376","https://openalex.org/W4416184858","https://openalex.org/W4417080477"],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"incur":[4],"significant":[5],"computational":[6],"and":[7,44,103,142],"memory":[8],"costs":[9],"when":[10],"processing":[11],"long":[12],"prompts,":[13],"as":[14],"full":[15],"self-attention":[16],"scales":[17],"quadratically":[18],"with":[19,127,147],"input":[20,124],"length.Token":[21],"compression":[22,60],"aims":[23],"to":[24,122],"address":[25],"this":[26,52],"challenge":[27],"by":[28,83,140,145],"reducing":[29],"the":[30,48,91,113],"number":[31],"of":[32,67,116],"tokens":[33],"representing":[34],"inputs.However,":[35],"existing":[36],"promptcompression":[37],"approaches":[38],"primarily":[39],"operate":[40],"in":[41,47,90,153],"token":[42,69],"space":[43],"overlook":[45],"inefficiencies":[46],"latent":[49],"embedding":[50,74],"space.In":[51],"paper,":[53],"we":[54],"propose":[55],"K-Token":[56,109],"Merging,":[57],"a":[58,72,76,84,131,149],"latent-space":[59],"framework":[61],"that":[62,108],"merges":[63],"each":[64],"contiguous":[65],"block":[66],"K":[68],"embeddings":[70],"into":[71],"single":[73],"via":[75],"lightweight":[77],"encoder.The":[78],"compressed":[79],"sequence":[80],"is":[81,155],"processed":[82],"LoRAadapted":[85],"LLM,":[86],"while":[87],"generation":[88],"remains":[89],"original":[92],"vocabulary.Experiments":[93],"on":[94,112],"structural":[95],"reasoning":[96],"(Textualized":[97],"Tree),":[98],"sentiment":[99],"classification":[100],"(Amazon":[101],"Reviews),":[102],"code":[104],"editing":[105],"(CommitPackFT)":[106],"show":[107],"Merging":[110,137],"lies":[111],"Pareto":[114],"frontier":[115],"performance":[117,129],"vs.":[118],"compression,":[119],"achieving":[120],"up":[121],"75%":[123],"length":[125],"reduction":[126],"minimal":[128],"degradation.On":[130],"Qwen":[132],"Coder":[133],"7B":[134],"model,":[135],"2-Token":[136],"reduces":[138],"latency":[139],"20.9%":[141],"improves":[143],"throughput":[144],"27.8%":[146],"only":[148],"1.1%":[150],"relative":[151],"increase":[152],"perplexity.Code":[154],"available":[156],"at":[157],"https://github.com/shsjxzh/":[158],"K-Token-Merging.":[159]},"counts_by_year":[],"updated_date":"2026-10-31T12:09:11.096373","created_date":"2026-04-18T00:00:00"}