{"id":"https://openalex.org/W4415098413","doi":"https://doi.org/10.48550/arxiv.2510.07318","title":"Artificial Hippocampus Networks for Efficient Long-Context Modeling","display_name":"Artificial Hippocampus Networks for Efficient Long-Context Modeling","publication_year":2025,"publication_date":"2025-10-08","ids":{"openalex":"https://openalex.org/W4415098413","doi":"https://doi.org/10.48550/arxiv.2510.07318"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2510.07318","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2510.07318","pdf_url":"https://arxiv.org/pdf/2510.07318","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/2510.07318","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101276796","display_name":"Yunhao Fang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fang, Yunhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Yu, Weihao","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yu, Weihao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100935761","display_name":"Zhong Shu","orcid":"https://orcid.org/0009-0003-5566-5608","observed_orcids":["https://orcid.org/0009-0003-5566-5608"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhong, Shu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5005965903","display_name":"Qinghao Ye","orcid":"https://orcid.org/0000-0002-7977-5540","observed_orcids":["https://orcid.org/0000-0002-7977-5540"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ye, Qinghao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5111029128","display_name":"Xuehan Xiong","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiong, Xuehan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":null,"display_name":"Wei, Lai","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wei, Lai","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":true,"cited_by_count":12,"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":0.8374000191688538,"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":0.8374000191688538,"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":0.0560000017285347,"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":0.027899999171495438,"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/long-context-modeling","display_name":"long-context modeling","score":0.9679999947547913},{"id":"https://openalex.org/keywords/sliding-window-attention","display_name":"sliding window attention","score":0.7990000247955322},{"id":"https://openalex.org/keywords/self-distillation","display_name":"self-distillation","score":0.6800000071525574},{"id":"https://openalex.org/keywords/mamba-2","display_name":"Mamba-2","score":0.6890000104904175},{"id":"https://openalex.org/keywords/open-weight-llms","display_name":"open-weight LLMs","score":0.6060000061988831},{"id":"https://openalex.org/keywords/inference-efficiency","display_name":"inference efficiency","score":0.5479999780654907}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8179000020027161},{"id":"https://openalex.org/C81081738","wikidata":"https://www.wikidata.org/wiki/Q55542","display_name":"Lossless compression","level":3,"score":0.6516000032424927},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.5065000057220459},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5047000050544739},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.4903999865055084},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.48660001158714294},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.435699999332428},{"id":"https://openalex.org/C102392041","wikidata":"https://www.wikidata.org/wiki/Q592860","display_name":"Sliding window protocol","level":3,"score":0.4307999908924103},{"id":"https://openalex.org/C115537543","wikidata":"https://www.wikidata.org/wiki/Q165596","display_name":"Cache","level":2,"score":0.40849998593330383},{"id":"https://openalex.org/C2776459999","wikidata":"https://www.wikidata.org/wiki/Q2119376","display_name":"Fidelity","level":2,"score":0.3855000138282776},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.3598000109195709},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.3319000005722046},{"id":"https://openalex.org/C2778751112","wikidata":"https://www.wikidata.org/wiki/Q835016","display_name":"Window (computing)","level":2,"score":0.3197999894618988},{"id":"https://openalex.org/C190502265","wikidata":"https://www.wikidata.org/wiki/Q17069496","display_name":"MNIST database","level":3,"score":0.3028999865055084},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.30239999294281006},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.29339998960494995},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.28060001134872437},{"id":"https://openalex.org/C176649486","wikidata":"https://www.wikidata.org/wiki/Q2308807","display_name":"Memory management","level":3,"score":0.2632000148296356}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2510.07318","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2510.07318","pdf_url":"https://arxiv.org/pdf/2510.07318","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.2510.07318","doi":"https://doi.org/10.48550/arxiv.2510.07318","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2510.07318","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:2510.07318","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2510.07318","pdf_url":"https://arxiv.org/pdf/2510.07318","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/W4415098413.pdf","grobid_xml":"https://content.openalex.org/works/W4415098413.grobid-xml"},"referenced_works_count":27,"referenced_works":["https://openalex.org/W1821462560","https://openalex.org/W2167839676","https://openalex.org/W2940744433","https://openalex.org/W2988394319","https://openalex.org/W4240553276","https://openalex.org/W4327810158","https://openalex.org/W4377865076","https://openalex.org/W4384648484","https://openalex.org/W4391709268","https://openalex.org/W4393399080","https://openalex.org/W4394655428","https://openalex.org/W4394737116","https://openalex.org/W4396815229","https://openalex.org/W4399414358","https://openalex.org/W4399598242","https://openalex.org/W4400611829","https://openalex.org/W4402667093","https://openalex.org/W4403571350","https://openalex.org/W4404330874","https://openalex.org/W4405655184","https://openalex.org/W4405766390","https://openalex.org/W4405902850","https://openalex.org/W4406071672","https://openalex.org/W4406465766","https://openalex.org/W4412886690","https://openalex.org/W4414281281","https://openalex.org/W4417054624"],"related_works":["https://openalex.org/W7129410104","https://openalex.org/W7222211659","https://openalex.org/W3093260394","https://openalex.org/W4404783458","https://openalex.org/W6891983197","https://openalex.org/W4379087052","https://openalex.org/W7127407113","https://openalex.org/W7119188126","https://openalex.org/W7214521629","https://openalex.org/W4396881792"],"abstract_inverted_index":{"Long-sequence":[0],"modeling":[1],"faces":[2],"a":[3,36,46,59,72,126],"fundamental":[4],"trade-off":[5],"between":[6],"the":[7,17,28,50,106,115,141,145,149,196],"efficiency":[8],"of":[9,19,39,49,132,152],"compressive":[10],"fixed-size":[11,73],"memory":[12,22,37,191,206],"in":[13,23,31],"RNN-like":[14,86],"models":[15,170],"and":[16,91,113,136,165,176,190,205],"fidelity":[18],"lossless":[20,55],"growing":[21],"attention-based":[24],"Transformers.":[25],"Inspired":[26],"by":[27,203,208],"Multi-Store":[29],"Model":[30],"cognitive":[32],"science,":[33],"we":[34,81],"introduce":[35],"framework":[38],"artificial":[40],"neural":[41],"networks.":[42],"Our":[43],"method":[44,104,124],"maintains":[45],"sliding":[47,129,173],"window":[48,130,174],"Transformer's":[51],"KV":[52],"cache":[53,207],"as":[54],"short-term":[56],"memory,":[57],"while":[58,186,210],"learnable":[60],"module":[61],"termed":[62],"Artificial":[63],"Hippocampus":[64],"Network":[65],"(AHN)":[66],"recurrently":[67],"compresses":[68],"out-of-window":[69],"information":[70],"into":[71],"compact":[74],"long-term":[75],"memory.":[76],"To":[77],"validate":[78],"this":[79],"framework,":[80],"instantiate":[82],"AHNs":[83,118,137,199],"using":[84],"modern":[85],"architectures,":[87],"including":[88],"Mamba2,":[89],"DeltaNet,":[90],"GatedDeltaNet":[92],"to":[93,183,222],"augment":[94],"open-weight":[95],"LLMs.":[96],"We":[97],"also":[98],"propose":[99],"an":[100],"efficient":[101],"self-distillation":[102],"training":[103],"where":[105],"base":[107],"model's":[108],"all":[109],"parameters":[110,116],"are":[111,119],"frozen":[112],"only":[114,139],"from":[117,220],"optimized.":[120],"For":[121,193],"inference,":[122],"our":[123],"sets":[125],"default":[127],"large":[128],"size":[131],"32k":[133,146],"for":[134],"attention,":[135],"activate":[138],"when":[140],"sequence":[142,218],"length":[143],"exceeds":[144],"window,":[147],"addressing":[148],"quadratic-complexity":[150],"issue":[151],"attention":[153],"that":[154,157,168],"emerges":[155],"at":[156],"scale.":[158],"Extensive":[159],"experiments":[160],"on":[161,215],"long-context":[162],"benchmarks":[163],"LV-Eval":[164,216],"InfiniteBench":[166],"demonstrate":[167],"AHN-augmented":[169],"consistently":[171],"outperform":[172],"baselines":[175],"achieve":[177],"performance":[178],"comparable":[179],"or":[180],"even":[181],"superior":[182],"full-attention":[184],"models,":[185],"substantially":[187],"reducing":[188],"computational":[189],"requirements.":[192],"instance,":[194],"augmenting":[195],"Qwen2.5-3B-Instruct":[197],"with":[198],"reduces":[200],"inference":[201],"FLOPs":[202],"40.5%":[204],"74.0%,":[209],"improving":[211],"its":[212],"average":[213],"score":[214],"(128k":[217],"length)":[219],"4.41":[221],"5.88.":[223],"Code":[224],"is":[225],"available":[226],"at:":[227],"https://github.com/ByteDance-Seed/AHN.":[228]},"counts_by_year":[{"year":2026,"cited_by_count":10},{"year":2025,"cited_by_count":2}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2025-10-13T00:00:00"}