{"id":"https://openalex.org/W7163210473","doi":"https://doi.org/10.48550/arxiv.2606.01412","title":"GPTQ-intrinsic LoRA: A Near-optimal Algorithm for Low-precision Quantization with Low-rank Adaptation","display_name":"GPTQ-intrinsic LoRA: A Near-optimal Algorithm for Low-precision Quantization with Low-rank Adaptation","publication_year":2026,"publication_date":"2026-05-31","ids":{"openalex":"https://openalex.org/W7163210473","doi":"https://doi.org/10.48550/arxiv.2606.01412"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.01412","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.01412","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.2606.01412","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137702373","display_name":"Shihao Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Shihao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5137653179","display_name":"Rayan Saab","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Saab, Rayan","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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.3043999969959259,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.3043999969959259,"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":0.27090001106262207,"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/T12535","display_name":"Machine Learning and Data Classification","score":0.039000000804662704,"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/quantization","display_name":"Quantization (signal processing)","score":0.7379999756813049},{"id":"https://openalex.org/keywords/bounded-function","display_name":"Bounded function","score":0.5449000000953674},{"id":"https://openalex.org/keywords/residual","display_name":"Residual","score":0.525600016117096},{"id":"https://openalex.org/keywords/regularization","display_name":"Regularization (linguistics)","score":0.4578999876976013},{"id":"https://openalex.org/keywords/vector-quantization","display_name":"Vector quantization","score":0.3917999863624573},{"id":"https://openalex.org/keywords/upper-and-lower-bounds","display_name":"Upper and lower bounds","score":0.3269999921321869},{"id":"https://openalex.org/keywords/artificial-neural-network","display_name":"Artificial neural network","score":0.32030001282691956}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.7379999756813049},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.6686999797821045},{"id":"https://openalex.org/C34388435","wikidata":"https://www.wikidata.org/wiki/Q2267362","display_name":"Bounded function","level":2,"score":0.5449000000953674},{"id":"https://openalex.org/C155512373","wikidata":"https://www.wikidata.org/wiki/Q287450","display_name":"Residual","level":2,"score":0.525600016117096},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.5209000110626221},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.4578999876976013},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4081000089645386},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.3917999863624573},{"id":"https://openalex.org/C77553402","wikidata":"https://www.wikidata.org/wiki/Q13222579","display_name":"Upper and lower bounds","level":2,"score":0.3269999921321869},{"id":"https://openalex.org/C50644808","wikidata":"https://www.wikidata.org/wiki/Q192776","display_name":"Artificial neural network","level":2,"score":0.32030001282691956},{"id":"https://openalex.org/C93372532","wikidata":"https://www.wikidata.org/wiki/Q6552455","display_name":"Linde\u2013Buzo\u2013Gray algorithm","level":3,"score":0.3059000074863434},{"id":"https://openalex.org/C2776359362","wikidata":"https://www.wikidata.org/wiki/Q2145286","display_name":"Representation (politics)","level":3,"score":0.2971999943256378},{"id":"https://openalex.org/C141379421","wikidata":"https://www.wikidata.org/wiki/Q6094427","display_name":"Iterative reconstruction","level":2,"score":0.2583000063896179},{"id":"https://openalex.org/C66322947","wikidata":"https://www.wikidata.org/wiki/Q11658","display_name":"Transformer","level":3,"score":0.2547000050544739},{"id":"https://openalex.org/C147764199","wikidata":"https://www.wikidata.org/wiki/Q6865248","display_name":"Minification","level":2,"score":0.25459998846054077},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.25360000133514404},{"id":"https://openalex.org/C175291020","wikidata":"https://www.wikidata.org/wiki/Q1156822","display_name":"Offset (computer science)","level":2,"score":0.2515999972820282}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.01412","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.01412","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.2606.01412","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.01412","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":{"Post-training":[0],"quantization":[1,13,99,173],"is":[2,22,58,135],"widely":[3],"used":[4],"for":[5,72],"compressing":[6],"large":[7],"neural":[8],"networks,":[9],"but":[10],"aggressive":[11],"low-bit":[12],"can":[14,177],"significantly":[15],"degrade":[16],"model":[17],"quality.":[18],"A":[19],"common":[20],"remedy":[21],"to":[23,33,64,143,162],"augment":[24],"the":[25,36,51,67,92,104,108,114,129,138,153],"quantized":[26],"weights":[27],"with":[28,180,184,212],"a":[29,59,87,97,171],"low-rank":[30,48,79,93,182,210],"correction,":[31],"leading":[32],"approximations":[34],"of":[35,119],"form":[37],"$W\\approx":[38],"Q+LR$.":[39],"In":[40],"this":[41,45,73],"paper,":[42],"we":[43,121],"study":[44],"low-precision":[46],"plus":[47],"representation":[49],"through":[50],"layer-wise":[52,123,187],"reconstruction":[53,124,188],"objective":[54],"$\\|XW-X(Q+LR)\\|_F^2$,":[55],"where":[56,111],"$X$":[57],"calibration":[60,105],"matrix.":[61],"We":[62,82,167],"establish,":[63],"our":[65],"knowledge,":[66],"first":[68],"information-theoretic":[69,154],"lower":[70,155],"bounds":[71,126,151,156],"problem":[74],"under":[75],"finite-alphabet":[76],"and":[77,164,195,206],"bounded":[78],"compensation":[80,183],"constraints.":[81],"then":[83],"propose":[84],"GPTQ-intrinsic":[85,201],"LoRA,":[86],"training-free":[88],"algorithm":[89],"that":[90,176,200],"incorporates":[91],"correction":[94],"directly":[95],"into":[96],"GPTQ-style":[98],"pass":[100],"by":[101,137,209],"appropriately":[102],"augmenting":[103],"Hessian.":[106],"For":[107],"choice":[109],"$L=V_r$,":[110],"$V_r$":[112],"contains":[113],"top":[115],"right":[116],"singular":[117],"vectors":[118],"$X$,":[120],"prove":[122],"error":[125],"in":[127,157],"which":[128],"usual":[130],"GPTQ":[131,205,207],"dependence":[132],"on":[133,191],"$\\|X\\|_F^2$":[134],"replaced":[136],"rank-$r$":[139],"residual":[140],"$\\|X-X_r\\|_F^2$,":[141],"up":[142,161],"regularization":[144],"terms.":[145],"Under":[146],"natural":[147],"structural":[148],"assumptions,":[149],"these":[150],"match":[152],"their":[158],"dominant":[159],"scaling,":[160],"constants":[163],"mild":[165],"factors.":[166],"also":[168],"introduce":[169],"Bid-Up,":[170],"fixed-grid":[172],"refinement":[174,216],"step":[175],"be":[178],"alternated":[179],"optimal":[181],"guaranteed":[185],"non-increasing":[186],"error.":[189],"Experiments":[190],"Qwen3":[192],"language":[193],"models":[194],"DeiT":[196],"vision":[197],"transformers":[198],"show":[199],"LoRA":[202],"improves":[203],"over":[204],"followed":[208],"compensation,":[211],"additional":[213],"gains":[214],"from":[215],"loops.":[217]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-03T00:00:00"}