{"id":"https://openalex.org/W7162288709","doi":"https://doi.org/10.48550/arxiv.2605.23464","title":"Unextractable Protocol Models: Collaborative Training and Inference without Weight Materialization","display_name":"Unextractable Protocol Models: Collaborative Training and Inference without Weight Materialization","publication_year":2026,"publication_date":"2026-05-22","ids":{"openalex":"https://openalex.org/W7162288709","doi":"https://doi.org/10.48550/arxiv.2605.23464"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.23464","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.23464","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.2605.23464","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136901627","display_name":"Alexander Long","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Long, Alexander","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5075325374","display_name":"Chamin Hewa Koneputugodage","orcid":"https://orcid.org/0009-0000-1333-5967"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Koneputugodage, Chamin Hewa","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5039055995","display_name":"Thalaiyasingam Ajanthan","orcid":"https://orcid.org/0000-0002-6431-0775"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ajanthan, Thalaiyasingam","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101436506","display_name":"Yan Zuo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zuo, Yan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136948014","display_name":"Gil Avraham","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Avraham, Gil","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5015805464","display_name":"Violetta Shevchenko","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shevchenko, Violetta","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5024380002","display_name":"Hadi M. Dolatabadi","orcid":"https://orcid.org/0000-0001-9418-1487"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dolatabadi, Hadi Mohaghegh","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136968103","display_name":"Sameera Ramasinghe","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ramasinghe, Sameera","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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.5275999903678894,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.5275999903678894,"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/T10764","display_name":"Privacy-Preserving Technologies in Data","score":0.23280000686645508,"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/T11598","display_name":"Internet Traffic Analysis and Secure E-voting","score":0.04490000009536743,"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/inference","display_name":"Inference","score":0.6420999765396118},{"id":"https://openalex.org/keywords/protocol","display_name":"Protocol (science)","score":0.5331000089645386},{"id":"https://openalex.org/keywords/rendering","display_name":"Rendering (computer graphics)","score":0.47749999165534973},{"id":"https://openalex.org/keywords/overhead","display_name":"Overhead (engineering)","score":0.46389999985694885},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.44920000433921814},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.4374000132083893}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7526000142097473},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.6420999765396118},{"id":"https://openalex.org/C2780385302","wikidata":"https://www.wikidata.org/wiki/Q367158","display_name":"Protocol (science)","level":3,"score":0.5331000089645386},{"id":"https://openalex.org/C205711294","wikidata":"https://www.wikidata.org/wiki/Q176953","display_name":"Rendering (computer graphics)","level":2,"score":0.47749999165534973},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.47269999980926514},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.46389999985694885},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.44920000433921814},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.4374000132083893},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.38359999656677246},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3407000005245209},{"id":"https://openalex.org/C14036430","wikidata":"https://www.wikidata.org/wiki/Q3736076","display_name":"Function (biology)","level":2,"score":0.336899995803833},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.33219999074935913},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.30550000071525574},{"id":"https://openalex.org/C2775924081","wikidata":"https://www.wikidata.org/wiki/Q55608371","display_name":"Control (management)","level":2,"score":0.2955000102519989},{"id":"https://openalex.org/C2984842247","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep neural networks","level":3,"score":0.2687000036239624},{"id":"https://openalex.org/C29122968","wikidata":"https://www.wikidata.org/wiki/Q1414816","display_name":"Incentive","level":2,"score":0.2662000060081482}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.23464","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.23464","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.2605.23464","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.23464","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":{"We":[0,50,159],"consider":[1,160],"a":[2,13,23,39,56,133],"decentralized":[3,216],"setup":[4,66],"in":[5,214],"which":[6],"the":[7,26,33,63,93,165,188],"participants":[8,75],"collaboratively":[9,200],"train":[10,192],"and":[11,17,58,104,121,142,155,172,177],"serve":[12],"large":[14],"neural":[15],"network,":[16],"where":[18,38],"each":[19],"participant":[20,90],"only":[21],"processes":[22],"subset":[24],"of":[25,35,167,181,187],"model.":[27],"In":[28],"this":[29],"setup,":[30],"we":[31,122],"explore":[32],"possibility":[34],"unmaterializable":[36],"weights,":[37],"full":[40],"weight":[41],"set":[42],"is":[43],"never":[44],"available":[45],"to":[46,67,125,152,174,191,198,209],"any":[47],"one":[48],"participant.":[49],"introduce":[51],"Unextractable":[52],"Protocol":[53],"Models":[54],"(UPMs):":[55],"training":[57,149],"inference":[59],"framework":[60],"that":[61,164,178],"leverages":[62],"sharded":[64],"model":[65,69],"ensure":[68],"shards":[70],"(i.e.,":[71],"subsets)":[72],"held":[73],"by":[74],"are":[76,170],"incompatible":[77],"at":[78,89,146],"different":[79],"time":[80,154],"steps.":[81],"UPMs":[82,205],"periodically":[83],"inject":[84],"time-varying,":[85],"random,":[86],"invertible":[87],"transforms":[88,107],"boundaries;":[91],"preserving":[92],"overall":[94],"network":[95],"function":[96],"yet":[97,202],"rendering":[98],"cross-time":[99],"assemblies":[100],"incoherent.":[101],"On":[102],"Qwen-2.5-0.5B":[103],"Llama-3.2-1B,":[105],"10,000":[106],"leave":[108],"FP32":[109],"perplexity":[110],"unchanged":[111],"($\u0394$PPL":[112],"$<":[113,117,156],"0.01$;":[114],"Jensen-Shannon":[115],"drift":[116],"4":[118],"\\times":[119],"10^{-5}$),":[120],"show":[123],"how":[124],"control":[126],"growth":[127],"for":[128],"lower":[129],"precision":[130],"datatypes.":[131],"Applying":[132],"transform":[134],"every":[135],"30s":[136],"adds":[137],"3%":[138],"latency,":[139],"0.1%":[140],"bandwidth,":[141],"10%":[143],"GPU-memory":[144],"overhead":[145,150],"inference,":[147],"while":[148],"falls":[151],"1.6%":[153],"1$%":[157],"memory.":[158],"several":[161],"attacks,":[162],"showing":[163],"requirements":[166],"direct":[168],"attacks":[169],"impractical":[171],"easy":[173],"defend":[175],"against,":[176],"gradient-based":[179],"fine-tuning":[180],"stitched":[182],"partitions":[183],"consumes":[184],"$\\geq":[185],"60$%":[186],"tokens":[189],"required":[190],"from":[193],"scratch.":[194],"By":[195],"enabling":[196],"models":[197],"be":[199],"trained":[201],"not":[203],"extracted,":[204],"make":[206],"it":[207],"practical":[208],"embed":[210],"programmatic":[211],"incentive":[212],"mechanisms":[213],"community-driven":[215],"training.":[217]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-26T00:00:00"}