{"id":"https://openalex.org/W7160614221","doi":"https://doi.org/10.48550/arxiv.2605.05693","title":"Saliency-Aware Regularized Quantization Calibration for Large Language Models","display_name":"Saliency-Aware Regularized Quantization Calibration for Large Language Models","publication_year":2026,"publication_date":"2026-05-07","ids":{"openalex":"https://openalex.org/W7160614221","doi":"https://doi.org/10.48550/arxiv.2605.05693"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.05693","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.05693","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,"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.2605.05693","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5081653312","display_name":"Yanlong Zhao","orcid":"https://orcid.org/0000-0001-8913-6032"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Yanlong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135643570","display_name":"Xiaoyuan Cheng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Cheng, Xiaoyuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5027749962","display_name":"Huihang Liu","orcid":"https://orcid.org/0000-0002-6195-7573"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Huihang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135648397","display_name":"Baihua He","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"He, Baihua","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135694496","display_name":"Xinyu Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Xinyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101564932","display_name":"Hong Zhu","orcid":"https://orcid.org/0000-0002-3232-1314"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhu, Harrison Bo Hua","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135643927","display_name":"Wenlong Chen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Chen, Wenlong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5126972369","display_name":"Li Zeng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zeng, Li","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5135662458","display_name":"Zhuo Sun","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sun, Zhuo","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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.1768999993801117,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.1768999993801117,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.14890000224113464,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.10189999639987946,"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":0.7279999852180481},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.5468000173568726},{"id":"https://openalex.org/keywords/regularization","display_name":"Regularization (linguistics)","score":0.45809999108314514},{"id":"https://openalex.org/keywords/perplexity","display_name":"Perplexity","score":0.36090001463890076},{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.2896000146865845}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.7279999852180481},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.5468000173568726},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4781000018119812},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.4715000092983246},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.462799996137619},{"id":"https://openalex.org/C2776135515","wikidata":"https://www.wikidata.org/wiki/Q17143721","display_name":"Regularization (linguistics)","level":2,"score":0.45809999108314514},{"id":"https://openalex.org/C100279451","wikidata":"https://www.wikidata.org/wiki/Q372193","display_name":"Perplexity","level":3,"score":0.36090001463890076},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.351500004529953},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3018999993801117},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.29649999737739563},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.2896000146865845},{"id":"https://openalex.org/C167085575","wikidata":"https://www.wikidata.org/wiki/Q6803654","display_name":"Mean squared prediction error","level":2,"score":0.2624000012874603},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.25699999928474426}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.05693","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.05693","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,"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.2605.05693","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.05693","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,"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":[1,23],"(PTQ)":[2],"is":[3],"an":[4],"effective":[5],"approach":[6],"for":[7],"deploying":[8],"large":[9],"language":[10],"models":[11],"(LLMs)":[12],"under":[13,166],"memory":[14],"and":[15,159,163,174,182],"latency":[16],"constraints.":[17],"Most":[18],"existing":[19,52,156],"PTQ":[20,53,99,157],"methods":[21,165],"determine":[22],"parameters":[24],"by":[25],"minimizing":[26],"a":[27,32,93,102,120,167],"layer-wise":[28],"reconstruction":[29,60],"error":[30,61],"on":[31,58,172],"predetermined":[33],"calibration":[34,54,66],"dataset,":[35],"typically":[36],"optimized":[37],"via":[38],"either":[39],"scale":[40],"search":[41],"or":[42,64],"Gram-based":[43,164],"methods.":[44],"However,":[45],"from":[46,74,109],"the":[47,70,75,110,140],"perspective":[48],"of":[49],"generalization":[50,148],"risk,":[51],"objectives":[55,100],"based":[56],"solely":[57],"empirical":[59],"over":[62],"limited":[63],"unrepresentative":[65],"data":[67],"may":[68],"move":[69],"quantized":[71,134],"weights":[72,135,142],"away":[73],"original":[76,111,141],"floating-point":[77],"weights,":[78],"potentially":[79],"degrading":[80],"downstream":[81],"performance.":[82],"To":[83],"address":[84],"this":[85,116],"issue,":[86],"we":[87],"propose":[88],"\\emph{Regularized":[89],"Quantization":[90,127],"Calibration}":[91,128],"(RQC),":[92],"unified":[94,168],"framework":[95,117],"that":[96,104],"augments":[97],"standard":[98],"with":[101],"regularizer":[103],"explicitly":[105],"controls":[106],"weight":[107],"deviation":[108],"weights.":[112],"We":[113],"further":[114],"generalize":[115],"to":[118,136,139,146],"incorporate":[119],"saliency-aware":[121],"regularizer,":[122],"resulting":[123],"in":[124,180],"\\emph{Saliency-Aware":[125],"Regularized":[126],"(SARQC).":[129],"The":[130],"proposed":[131],"regularization":[132],"encourages":[133],"remain":[137],"close":[138],"during":[143],"calibration,":[144],"leading":[145],"improved":[147],"at":[149],"inference":[150,188],"time.":[151],"SARQC":[152],"integrates":[153],"seamlessly":[154],"into":[155],"pipelines":[158],"enhances":[160],"both":[161],"scale-search-based":[162],"formulation.":[169],"Extensive":[170],"experiments":[171],"dense":[173],"Mixture-of-Experts":[175],"LLMs":[176],"demonstrate":[177],"consistent":[178],"improvements":[179],"perplexity":[181],"zero-shot":[183],"accuracy,":[184],"without":[185],"introducing":[186],"additional":[187],"overhead.":[189]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-05-09T00:00:00"}