{"id":"https://openalex.org/W7117581310","doi":"https://doi.org/10.48550/arxiv.2512.21651","title":"Rethinking Output Alignment For 1-bit Post-Training Quantization of Large Language Models","display_name":"Rethinking Output Alignment For 1-bit Post-Training Quantization of Large Language Models","publication_year":2025,"publication_date":"2025-12-25","ids":{"openalex":"https://openalex.org/W7117581310","doi":"https://doi.org/10.48550/arxiv.2512.21651"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2512.21651","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2512.21651","pdf_url":"https://arxiv.org/pdf/2512.21651","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","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2512.21651","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5103837931","display_name":"Dung Anh Hoang","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hoang, Dung Anh","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5121564103","display_name":"Cuong Pham","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I56590836","display_name":"Monash University","ror":"https://ror.org/02bfwt286","country_code":"AU","type":"education","lineage":["https://openalex.org/I56590836"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"Pham, Cuong","raw_affiliation_strings":["Department of Data Science and AI Monash University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Data Science and AI Monash University","institution_ids":["https://openalex.org/I56590836"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134963275","display_name":"Cuong Nguyen","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I28290843","display_name":"University of Surrey","ror":"https://ror.org/00ks66431","country_code":"GB","type":"education","lineage":["https://openalex.org/I28290843"]}],"countries":["GB"],"is_corresponding":false,"raw_author_name":"Nguyen, Cuong","raw_affiliation_strings":["Centre for Vision , Speech and Signal Processing University of Surrey"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Centre for Vision , Speech and Signal Processing University of Surrey","institution_ids":["https://openalex.org/I28290843"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5121578959","display_name":"Trung le","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I56590836","display_name":"Monash University","ror":"https://ror.org/02bfwt286","country_code":"AU","type":"education","lineage":["https://openalex.org/I56590836"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"le, Trung","raw_affiliation_strings":["Department of Data Science and AI Monash University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Data Science and AI Monash University","institution_ids":["https://openalex.org/I56590836"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5125181377","display_name":"Jianfei Cai","orcid":"https://orcid.org/0000-0002-9444-3763","observed_orcids":["https://orcid.org/0000-0002-9444-3763","https://orcid.org/0009-0000-8860-2767"]},"institutions":[{"id":"https://openalex.org/I56590836","display_name":"Monash University","ror":"https://ror.org/02bfwt286","country_code":"AU","type":"education","lineage":["https://openalex.org/I56590836"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"Cai, Jianfei","raw_affiliation_strings":["Department of Data Science and AI Monash University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Data Science and AI Monash University","institution_ids":["https://openalex.org/I56590836"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5025723803","display_name":"Thanh-Toan Do","orcid":"https://orcid.org/0000-0002-6249-0848","observed_orcids":["https://orcid.org/0000-0002-6249-0848"]},"institutions":[{"id":"https://openalex.org/I56590836","display_name":"Monash University","ror":"https://ror.org/02bfwt286","country_code":"AU","type":"education","lineage":["https://openalex.org/I56590836"]}],"countries":["AU"],"is_corresponding":false,"raw_author_name":"Do, Thanh-Toan","raw_affiliation_strings":["Department of Data Science and AI Monash University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Data Science and AI Monash University","institution_ids":["https://openalex.org/I56590836"]}]}],"institutions":[],"countries_distinct_count":2,"institutions_distinct_count":2,"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":0.9779999852180481,"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":0.9779999852180481,"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":0.0034000000450760126,"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/T10028","display_name":"Topic Modeling","score":0.0031999999191612005,"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/one-bit-quantization","display_name":"1-bit quantization","score":0.8650000095367432},{"id":"https://openalex.org/keywords/post-training-quantization","display_name":"post-training quantization","score":1.0},{"id":"https://openalex.org/keywords/large-language-models","display_name":"large language models","score":0.9829999804496765},{"id":"https://openalex.org/keywords/model-compression","display_name":"model compression","score":0.6549999713897705},{"id":"https://openalex.org/keywords/low-bit-neural-networks","display_name":"low-bit neural networks","score":0.41499999165534973}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.7473000288009644},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6463000178337097},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.48969998955726624},{"id":"https://openalex.org/C147764199","wikidata":"https://www.wikidata.org/wiki/Q6865248","display_name":"Minification","level":2,"score":0.3747999966144562},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.37220001220703125},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.35010001063346863},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.3118000030517578},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.30550000071525574},{"id":"https://openalex.org/C2778712577","wikidata":"https://www.wikidata.org/wiki/Q3505966","display_name":"Retraining","level":2,"score":0.2939999997615814},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.28279998898506165},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.273499995470047},{"id":"https://openalex.org/C105339364","wikidata":"https://www.wikidata.org/wiki/Q2297740","display_name":"Software deployment","level":2,"score":0.2721000015735626},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.26420000195503235},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.25360000133514404},{"id":"https://openalex.org/C2778671685","wikidata":"https://www.wikidata.org/wiki/Q219239","display_name":"Recipe","level":2,"score":0.25200000405311584}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2512.21651","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2512.21651","pdf_url":"https://arxiv.org/pdf/2512.21651","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.2512.21651","doi":"https://doi.org/10.48550/arxiv.2512.21651","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2512.21651","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:2512.21651","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2512.21651","pdf_url":"https://arxiv.org/pdf/2512.21651","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/W7117581310.pdf","grobid_xml":"https://content.openalex.org/works/W7117581310.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7212031868","https://openalex.org/W4416179544","https://openalex.org/W4391670689","https://openalex.org/W7214472452","https://openalex.org/W4407764019","https://openalex.org/W4409150022","https://openalex.org/W4396913356","https://openalex.org/W4387635190","https://openalex.org/W4414359705","https://openalex.org/W4412888573"],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"deliver":[4],"strong":[5],"performance":[6],"across":[7,182],"a":[8,60,110,199],"wide":[9],"range":[10],"of":[11,83,136,189],"NLP":[12],"tasks,":[13],"but":[14],"their":[15,25],"massive":[16],"sizes":[17],"hinder":[18],"deployment":[19],"on":[20,194],"resource-constrained":[21],"devices.":[22],"To":[23],"reduce":[24],"computational":[26,213],"and":[27,39,58],"memory":[28],"burden,":[29],"various":[30],"compression":[31],"techniques":[32],"have":[33,73],"been":[34],"proposed,":[35],"including":[36],"quantization,":[37],"pruning,":[38],"knowledge":[40],"distillation.":[41],"Among":[42],"these,":[43],"post-training":[44,71],"quantization":[45,72,90,99],"(PTQ)":[46],"is":[47,100,116,126,141],"widely":[48],"adopted":[49],"for":[50,63,70,203],"its":[51],"efficiency,":[52],"as":[53],"it":[54],"requires":[55],"no":[56],"retraining":[57],"only":[59],"small":[61],"dataset":[62],"calibration,":[64],"enabling":[65],"low-cost":[66],"deployment.":[67],"Recent":[68],"advances":[69],"demonstrated":[74],"that":[75,146,172,206,218],"even":[76,161],"near":[77],"4-bit":[78],"methods":[79],"can":[80],"maintain":[81],"most":[82],"the":[84,121,129,134,164,190],"original":[85],"model":[86,150],"performance.":[87],"However,":[88,113],"1-bit":[89,98,165,204,224],"remains":[91],"particularly":[92],"challenging.":[93],"A":[94,138],"common":[95],"strategy":[96],"in":[97,149,163],"to":[101,127,142],"determine":[102],"binary":[103],"weights":[104],"by":[105],"matching":[106],"full-precision":[107],"parameters,":[108],"following":[109],"weight-driven":[111],"criterion.":[112],"this":[114,168,173],"objective":[115],"not":[117],"directly":[118],"aligned":[119],"with":[120],"quantized":[122],"model's":[123,130],"objective,":[124],"which":[125],"preserve":[128],"output":[131],"behavior":[132],"under":[133],"impact":[135],"quantization.":[137],"natural":[139],"alternative":[140],"adopt":[143],"output-driven":[144,157],"criteria":[145],"minimize":[147],"discrepancies":[148],"outputs":[151],"using":[152],"calibration":[153],"data.":[154],"Surprisingly,":[155],"naive":[156],"approaches":[158],"often":[159],"perform":[160],"worse":[162],"regime.":[166],"In":[167],"paper,":[169],"we":[170,197],"show":[171],"failure":[174],"arises":[175],"from":[176],"two":[177],"fundamental":[178],"issues:":[179],"error":[180],"accumulation":[181],"layers":[183],"and,":[184],"more":[185],"critically,":[186],"\\emph{anisotropic":[187],"distortion}":[188],"representation":[191],"space.":[192],"Based":[193],"these":[195,209],"insights,":[196],"propose":[198],"novel":[200],"PTQ":[201,225],"method":[202],"LLMs":[205],"explicitly":[207],"addresses":[208],"issues":[210],"while":[211],"maintaining":[212],"efficiency.":[214],"Extensive":[215],"experiments":[216],"demonstrate":[217],"our":[219],"approach":[220],"consistently":[221],"outperforms":[222],"existing":[223],"methods.":[226]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2025-12-30T00:00:00"}