{"id":"https://openalex.org/W4416613151","doi":"https://doi.org/10.48550/arxiv.2505.23735","title":"ATLAS: Learning to Optimally Memorize the Context at Test Time","display_name":"ATLAS: Learning to Optimally Memorize the Context at Test Time","publication_year":2025,"publication_date":"2025-05-29","ids":{"openalex":"https://openalex.org/W4416613151","doi":"https://doi.org/10.48550/arxiv.2505.23735"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2505.23735","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2505.23735","pdf_url":"https://arxiv.org/pdf/2505.23735","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"},"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/2505.23735","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5066381651","display_name":"Ali Behrouz","orcid":"https://orcid.org/0000-0002-4934-669X","observed_orcids":["https://orcid.org/0000-0002-4934-669X"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Behrouz, Ali","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Li, Zeman","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Zeman","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5023530771","display_name":"Praneeth Kacham","orcid":"https://orcid.org/0000-0002-2207-4882","observed_orcids":["https://orcid.org/0000-0002-2207-4882"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kacham, Praneeth","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5064845272","display_name":"Majid Daliri","orcid":"https://orcid.org/0000-0003-4001-4346","observed_orcids":["https://orcid.org/0000-0003-4001-4346"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Daliri, Majid","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101718761","display_name":"Yuan Deng","orcid":"https://orcid.org/0000-0001-8425-6326","observed_orcids":["https://orcid.org/0000-0001-8425-6326"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Deng, Yuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102201220","display_name":"Peilin Zhong","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhong, Peilin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5041237539","display_name":"Meisam Razaviyayn","orcid":"https://orcid.org/0000-0003-4342-6661","observed_orcids":["https://orcid.org/0000-0003-4342-6661"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Razaviyayn, Meisam","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5075598023","display_name":"Vahab Mirrokni","orcid":"https://orcid.org/0000-0001-6705-5629","observed_orcids":["https://orcid.org/0000-0001-6705-5629"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mirrokni, Vahab","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":75,"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":0.9223999977111816,"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":0.9223999977111816,"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":0.029200000688433647,"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":0.009800000116229057,"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/in-context-learning","display_name":"in-context learning","score":0.628000020980835},{"id":"https://openalex.org/keywords/long-context-understanding","display_name":"long-context understanding","score":0.7990000247955322},{"id":"https://openalex.org/keywords/memory-capacity","display_name":"memory capacity","score":0.578000009059906},{"id":"https://openalex.org/keywords/linear-recurrent-neural-networks","display_name":"linear recurrent neural networks","score":0.5220000147819519},{"id":"https://openalex.org/keywords/language-modeling","display_name":"language modeling","score":0.6159999966621399},{"id":"https://openalex.org/keywords/commonsense-reasoning","display_name":"commonsense reasoning","score":0.9729999899864197},{"id":"https://openalex.org/keywords/long-context-retrieval","display_name":"long-context retrieval","score":0.4449999928474426},{"id":"https://openalex.org/keywords/atlas","display_name":"ATLAS","score":0.382999986410141}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7609000205993652},{"id":"https://openalex.org/C30038468","wikidata":"https://www.wikidata.org/wiki/Q4354775","display_name":"Memorization","level":2,"score":0.694100022315979},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5623999834060669},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5144000053405762},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.491100013256073},{"id":"https://openalex.org/C45340560","wikidata":"https://www.wikidata.org/wiki/Q215382","display_name":"Disjoint sets","level":2,"score":0.36469998955726624},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.328900009393692},{"id":"https://openalex.org/C12186640","wikidata":"https://www.wikidata.org/wiki/Q6815743","display_name":"Memory model","level":3,"score":0.3127000033855438},{"id":"https://openalex.org/C147168706","wikidata":"https://www.wikidata.org/wiki/Q1457734","display_name":"Recurrent neural network","level":3,"score":0.3111000061035156},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.30300000309944153},{"id":"https://openalex.org/C183322885","wikidata":"https://www.wikidata.org/wiki/Q17007702","display_name":"Context model","level":3,"score":0.29490000009536743},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.29179999232292175},{"id":"https://openalex.org/C2779602883","wikidata":"https://www.wikidata.org/wiki/Q15544750","display_name":"Memory architecture","level":2,"score":0.2879999876022339},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.2824999988079071},{"id":"https://openalex.org/C40506919","wikidata":"https://www.wikidata.org/wiki/Q7452469","display_name":"Sequence learning","level":2,"score":0.2818000018596649},{"id":"https://openalex.org/C34388435","wikidata":"https://www.wikidata.org/wiki/Q2267362","display_name":"Bounded function","level":2,"score":0.2669999897480011},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.2538999915122986},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.251800000667572}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2505.23735","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2505.23735","pdf_url":"https://arxiv.org/pdf/2505.23735","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.2505.23735","doi":"https://doi.org/10.48550/arxiv.2505.23735","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2505.23735","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":"pmh:oai:arXiv.org:2505.23735","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2505.23735","pdf_url":"https://arxiv.org/pdf/2505.23735","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"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416613151.pdf","grobid_xml":"https://content.openalex.org/works/W4416613151.grobid-xml"},"referenced_works_count":40,"referenced_works":["https://openalex.org/W1981530182","https://openalex.org/W2012903341","https://openalex.org/W2133564696","https://openalex.org/W2161278885","https://openalex.org/W2169976759","https://openalex.org/W2473344385","https://openalex.org/W2794325560","https://openalex.org/W2899663614","https://openalex.org/W2946609015","https://openalex.org/W2946659172","https://openalex.org/W2951365061","https://openalex.org/W2963748441","https://openalex.org/W3139537596","https://openalex.org/W3182042768","https://openalex.org/W4313447747","https://openalex.org/W4327810158","https://openalex.org/W4384648484","https://openalex.org/W4386721561","https://openalex.org/W4389539730","https://openalex.org/W4389599535","https://openalex.org/W4392085961","https://openalex.org/W4393213256","https://openalex.org/W4393247957","https://openalex.org/W4394708852","https://openalex.org/W4396881410","https://openalex.org/W4399655853","https://openalex.org/W4400435102","https://openalex.org/W4403849929","https://openalex.org/W4405253287","https://openalex.org/W4405254248","https://openalex.org/W4406032263","https://openalex.org/W4406755454","https://openalex.org/W4407569514","https://openalex.org/W4410412107","https://openalex.org/W4413146174","https://openalex.org/W4415797189","https://openalex.org/W4417091806","https://openalex.org/W6947991961","https://openalex.org/W6967103305","https://openalex.org/W7196932094"],"related_works":["https://openalex.org/W4406032263","https://openalex.org/W7222211659","https://openalex.org/W7108662860","https://openalex.org/W4393160131","https://openalex.org/W4285272557","https://openalex.org/W4400517511","https://openalex.org/W7162039850","https://openalex.org/W7116175540","https://openalex.org/W3196756515","https://openalex.org/W4394867741"],"abstract_inverted_index":{"Transformers":[0,224],"have":[1],"been":[2],"established":[3],"as":[4,52],"the":[5,22,105,113,122,128,160,164,168,174,200,221,233],"most":[6],"popular":[7],"backbones":[8],"in":[9,17,38,66,72,94,242],"sequence":[10],"modeling,":[11,209],"mainly":[12],"due":[13],"to":[14,24,46,81,127,158],"their":[15,36,63,95,137],"effectiveness":[16],"in-context":[18],"retrieval":[19],"tasks":[20,73,216],"and":[21,31,41,79,109,131,170,213,225],"ability":[23],"learn":[25],"at":[26],"scale.":[27],"Their":[28],"quadratic":[29],"memory":[30,60,99,108,123,151,165,179],"time":[32],"complexity,":[33],"however,":[34],"bound":[35],"applicability":[37],"longer":[39,82],"sequences":[40],"so":[42],"has":[43],"motivated":[44],"researchers":[45],"explore":[47],"effective":[48],"alternative":[49],"architectures":[50],"such":[51],"modern":[53],"recurrent":[54,59,228],"neural":[55],"networks":[56],"(a.k.a":[57],"long-term":[58,150,178],"module).":[61],"Despite":[62],"recent":[64,226],"success":[65],"diverse":[67],"downstream":[68],"tasks,":[69],"they":[70],"struggle":[71],"that":[74,86,101,156,195,218],"requires":[75],"long":[76,234],"context":[77,161,235,244],"understanding":[78,215],"extrapolation":[80],"sequences.":[83],"We":[84],"observe":[85],"these":[87,143],"shortcomings":[88],"come":[89],"from":[90],"three":[91,144],"disjoint":[92],"aspects":[93],"design:":[96],"(1)":[97],"limited":[98],"capacity":[100,155],"is":[102],"bounded":[103],"by":[104,162],"architecture":[106],"of":[107,112,118,136,177,190,199,223,237,246],"feature":[110],"mapping":[111],"input;":[114,130],"(2)":[115],"online":[116,175],"nature":[117,176],"update,":[119],"i.e.,":[120],"optimizing":[121,163],"only":[124],"with":[125,153],"respect":[126],"last":[129],"(3)":[132],"less":[133],"expressive":[134],"management":[135],"fixed-size":[138],"memory.":[139],"To":[140],"enhance":[141],"all":[142],"aspects,":[145],"we":[146,185],"present":[147,186],"ATLAS,":[148],"a":[149,187],"module":[152],"high":[154],"learns":[157],"memorize":[159],"based":[166],"on":[167,182,207],"current":[169],"past":[171],"tokens,":[172],"overcoming":[173],"models.":[180,229],"Building":[181],"this":[183],"insight,":[184],"new":[188],"family":[189],"Transformer-like":[191],"architectures,":[192],"called":[193],"DeepTransformers,":[194],"are":[196],"strict":[197],"generalizations":[198],"original":[201],"Transformer":[202],"architecture.":[203],"Our":[204],"experimental":[205],"results":[206],"language":[208],"common-sense":[210],"reasoning,":[211],"recall-intensive,":[212],"long-context":[214],"show":[217],"ATLAS":[219,230],"surpasses":[220],"performance":[222,236],"linear":[227],"further":[231],"improves":[232],"Titans,":[238],"achieving":[239],"+80\\%":[240],"accuracy":[241],"10M":[243],"length":[245],"BABILong":[247],"benchmark.":[248]},"counts_by_year":[{"year":2026,"cited_by_count":53},{"year":2025,"cited_by_count":21},{"year":2024,"cited_by_count":1}],"updated_date":"2026-11-10T11:39:43.021295","created_date":"2025-10-10T00:00:00"}