{"id":"https://openalex.org/W4414977014","doi":"https://doi.org/10.48550/arxiv.2510.05468","title":"AMAQ: Adaptive Mixed-bit Activation Quantization for Collaborative Parameter Efficient Fine-tuning","display_name":"AMAQ: Adaptive Mixed-bit Activation Quantization for Collaborative Parameter Efficient Fine-tuning","publication_year":2025,"publication_date":"2025-10-07","ids":{"openalex":"https://openalex.org/W4414977014","doi":"https://doi.org/10.48550/arxiv.2510.05468"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2510.05468","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2510.05468","pdf_url":"https://arxiv.org/pdf/2510.05468","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","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2510.05468","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5082442045","display_name":"Yurun Song","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Song, Yurun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5023011245","display_name":"Zhuoyi Yang","orcid":"https://orcid.org/0009-0008-7909-4930","observed_orcids":["https://orcid.org/0009-0008-7909-4930"]},"institutions":[{"id":"https://openalex.org/I204250578","display_name":"University of California, Irvine","ror":"https://ror.org/04gyf1771","country_code":"US","type":"education","lineage":["https://openalex.org/I204250578"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yang, Zhuoyi","raw_affiliation_strings":["UC Irvine"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UC Irvine","institution_ids":["https://openalex.org/I204250578"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5037855756","display_name":"Ian G. Harris","orcid":"https://orcid.org/0000-0003-3244-6528","observed_orcids":["https://orcid.org/0000-0003-3244-6528"]},"institutions":[{"id":"https://openalex.org/I204250578","display_name":"University of California, Irvine","ror":"https://ror.org/04gyf1771","country_code":"US","type":"education","lineage":["https://openalex.org/I204250578"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Harris, Ian G.","raw_affiliation_strings":["UC Irvine"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"UC Irvine","institution_ids":["https://openalex.org/I204250578"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5058967466","display_name":"Sangeetha Abdu Jyothi","orcid":"https://orcid.org/0009-0000-0503-4478","observed_orcids":["https://orcid.org/0009-0000-0503-4478"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jyothi, Sangeetha Abdu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.8906000256538391,"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"}},"topics":[{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.8906000256538391,"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"}},{"id":"https://openalex.org/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.0203000009059906,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.019500000402331352,"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/activation-quantization","display_name":"activation quantization","score":0.8970000147819519},{"id":"https://openalex.org/keywords/parameter-efficient-fine-tuning","display_name":"parameter-efficient fine-tuning","score":0.8009999990463257},{"id":"https://openalex.org/keywords/split-learning","display_name":"split learning","score":0.9049999713897705},{"id":"https://openalex.org/keywords/collaborative-training","display_name":"collaborative training","score":0.7160000205039978},{"id":"https://openalex.org/keywords/communication-efficiency","display_name":"communication efficiency","score":0.7860000133514404},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.7940000295639038},{"id":"https://openalex.org/keywords/training-stability","display_name":"training stability","score":0.5740000009536743}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.808899998664856},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.7229999899864197},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.5580000281333923},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5156999826431274},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.38940000534057617},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.3776000142097473},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3700000047683716},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.3547999858856201},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.3521000146865845},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.34869998693466187},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3425000011920929},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.336899995803833},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3122999966144562},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.3046000003814697},{"id":"https://openalex.org/C139807058","wikidata":"https://www.wikidata.org/wiki/Q352374","display_name":"Adaptation (eye)","level":2,"score":0.3012999892234802},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.2705000042915344},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2700999975204468},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.26249998807907104},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2612000107765198},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.26030001044273376},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.25679999589920044},{"id":"https://openalex.org/C157764524","wikidata":"https://www.wikidata.org/wiki/Q1383412","display_name":"Throughput","level":3,"score":0.2565000057220459},{"id":"https://openalex.org/C12269588","wikidata":"https://www.wikidata.org/wiki/Q132364","display_name":"Communications protocol","level":2,"score":0.25619998574256897}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2510.05468","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2510.05468","pdf_url":"https://arxiv.org/pdf/2510.05468","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.2510.05468","doi":"https://doi.org/10.48550/arxiv.2510.05468","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2510.05468","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:2510.05468","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2510.05468","pdf_url":"https://arxiv.org/pdf/2510.05468","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/W4414977014.pdf","grobid_xml":"https://content.openalex.org/works/W4414977014.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7138053792","https://openalex.org/W7077910239","https://openalex.org/W7222927907","https://openalex.org/W7131910536","https://openalex.org/W7166888950","https://openalex.org/W4407949540","https://openalex.org/W7163610126","https://openalex.org/W4401661003","https://openalex.org/W7162120183","https://openalex.org/W7103170887"],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"are":[4],"scaling":[5],"rapidly,":[6],"creating":[7],"significant":[8],"challenges":[9],"for":[10,40,125,171,186],"collaborative":[11,41,51,158,187],"server":[12],"client":[13],"distributed":[14],"training,":[15,52],"particularly":[16],"in":[17,50],"terms":[18],"of":[19],"communication":[20,48,169,191],"efficiency":[21,37],"and":[22,38,67,97,119,130,140,183],"computational":[23],"overheads.":[24],"To":[25,46],"address":[26],"these":[27],"challenges,":[28],"we":[29,53],"implement":[30],"Parameter-efficient":[31],"Split":[32],"Learning,":[33],"which":[34],"effectively":[35,87,154],"balances":[36],"performance":[39],"training":[42,138,159,188],"on":[43,94],"low-resource":[44],"devices.":[45],"reduce":[47],"overhead":[49,170],"introduce":[54],"Adaptive":[55],"Mixed":[56],"bit":[57,89,102,107,143],"Activation":[58],"Quantization":[59],"(AMAQ),":[60],"a":[61,167,181],"strategy":[62],"that":[63,151],"progressively":[64],"compresses":[65],"activations":[66],"gradients":[68],"from":[69],"high":[70],"precision":[71,78],"(6":[72],"to":[73,76,80],"8":[74],"bits)":[75],"low":[77],"(3":[79],"4":[81],"bits).":[82],"AMAQ":[83,109,152,180],"achieves":[84],"this":[85],"by":[86],"allocating":[88],"budgets":[90],"across":[91],"channels":[92],"based":[93],"feature":[95],"wise":[96,99],"layer":[98],"importance":[100],"using":[101],"regularization.":[103],"Under":[104],"the":[105,147],"same":[106],"budgets,":[108],"outperforms":[110],"fixed-precision":[111],"approaches,":[112],"delivering":[113],"about":[114,120],"2.5%":[115],"higher":[116],"generation":[117],"accuracy":[118,124,164],"1.3%":[121],"better":[122],"classification":[123],"models":[126],"like":[127],"LLaMA3":[128],"8B":[129],"Qwen2.5":[131],"7B.":[132],"In":[133],"addition,":[134],"it":[135],"significantly":[136],"enhances":[137],"stability":[139],"reducing":[141],"ultra-low":[142],"representation":[144],"collapse":[145],"during":[146,174],"training.":[148,175],"Experiments":[149],"demonstrate":[150],"integrates":[153],"into":[155],"practical":[156,182],"multi-machine":[157],"setups,":[160],"offering":[161],"superior":[162],"inference":[163],"with":[165,189],"only":[166],"modest":[168],"bits":[172],"adaptation":[173],"This":[176],"trade":[177],"off":[178],"makes":[179],"effective":[184],"solution":[185],"minimal":[190],"cost.":[192]},"counts_by_year":[],"updated_date":"2026-09-30T09:27:27.532063","created_date":"2025-10-10T00:00:00"}