{"id":"https://openalex.org/W7164035802","doi":"https://doi.org/10.48550/arxiv.2606.07819","title":"Joint Structural Pruning and Mixed-Precision Quantization for LLM Compression","display_name":"Joint Structural Pruning and Mixed-Precision Quantization for LLM Compression","publication_year":2026,"publication_date":"2026-06-05","ids":{"openalex":"https://openalex.org/W7164035802","doi":"https://doi.org/10.48550/arxiv.2606.07819"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.07819","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.07819","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":null,"license_id":null,"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.2606.07819","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5138239145","display_name":"Hoang-Loc La","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"La, Hoang-Loc","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128023380","display_name":"Truong-Thanh Le","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Le, Truong-Thanh","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138250090","display_name":"Amir Taherkordi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Taherkordi, Amir","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5029307415","display_name":"Phuong Hoai Ha","orcid":"https://orcid.org/0000-0001-8366-5590"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ha, Phuong Hoai","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/T10181","display_name":"Natural Language Processing Techniques","score":0.28119999170303345,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.28119999170303345,"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":0.20720000565052032,"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/T10901","display_name":"Advanced Data Compression Techniques","score":0.06509999930858612,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.7827000021934509},{"id":"https://openalex.org/keywords/perplexity","display_name":"Perplexity","score":0.7752000093460083},{"id":"https://openalex.org/keywords/memory-footprint","display_name":"Memory footprint","score":0.4681999981403351},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.45260000228881836},{"id":"https://openalex.org/keywords/vector-quantization","display_name":"Vector quantization","score":0.3495999872684479},{"id":"https://openalex.org/keywords/linde\u2013buzo\u2013gray-algorithm","display_name":"Linde\u2013Buzo\u2013Gray algorithm","score":0.31610000133514404}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.7827000021934509},{"id":"https://openalex.org/C100279451","wikidata":"https://www.wikidata.org/wiki/Q372193","display_name":"Perplexity","level":3,"score":0.7752000093460083},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6365000009536743},{"id":"https://openalex.org/C74912251","wikidata":"https://www.wikidata.org/wiki/Q6815727","display_name":"Memory footprint","level":2,"score":0.4681999981403351},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.45260000228881836},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4004000127315521},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.39259999990463257},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.3495999872684479},{"id":"https://openalex.org/C93372532","wikidata":"https://www.wikidata.org/wiki/Q6552455","display_name":"Linde\u2013Buzo\u2013Gray algorithm","level":3,"score":0.31610000133514404},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.3131999969482422},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.31299999356269836},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.29750001430511475},{"id":"https://openalex.org/C78548338","wikidata":"https://www.wikidata.org/wiki/Q2493","display_name":"Data compression","level":2,"score":0.296099990606308},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.2874999940395355},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.2599000036716461},{"id":"https://openalex.org/C193319292","wikidata":"https://www.wikidata.org/wiki/Q272172","display_name":"Hamming distance","level":2,"score":0.25870001316070557},{"id":"https://openalex.org/C2780069185","wikidata":"https://www.wikidata.org/wiki/Q7977945","display_name":"Equivalence (formal languages)","level":2,"score":0.25429999828338623}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.07819","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.07819","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2606.07819","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.07819","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":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"score":0.4231743812561035,"id":"https://metadata.un.org/sdg/4","display_name":"Quality Education"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Recently,":[0],"the":[1,52,104,184],"efficiency":[2],"of":[3],"Large":[4],"Language":[5],"Models":[6],"(LLMs)":[7],"deployment":[8],"has":[9],"become":[10],"a":[11,42,77,92,117,133],"critical":[12],"concern":[13],"in":[14,56,68,85],"practical":[15],"applications.":[16],"While":[17],"post-training":[18],"quantization":[19,39,67,130,147,161,166],"(PTQ)":[20],"and":[21,31,49,66,128,173,179,195],"structural":[22,125],"pruning":[23,65,126],"are":[24],"established":[25],"techniques":[26],"for":[27],"reducing":[28],"memory":[29],"footprint":[30],"inference":[32],"latency,":[33],"most":[34],"existing":[35],"PTQ":[36,95],"approaches":[37],"optimize":[38],"errors":[40,47],"on":[41,113,177],"per-layer":[43],"basis,":[44],"overlooking":[45],"how":[46],"accumulate":[48],"propagate":[50],"through":[51],"network,":[53],"often":[54],"resulting":[55],"suboptimal":[57],"solutions.":[58],"Traditional":[59],"pipelines":[60],"also":[61],"tend":[62],"to":[63,154,157,171,183],"apply":[64],"isolation":[69],"or":[70],"sequentially,":[71],"further":[72],"compounding":[73],"sub-optimality.":[74],"We":[75],"introduce":[76],"novel":[78,93,118],"end-to-end":[79],"framework":[80],"that":[81,97,122],"addresses":[82],"these":[83],"limitations":[84],"two":[86],"key":[87],"ways.":[88],"First,":[89],"we":[90,115],"propose":[91],"mixed-precision":[94,129],"strategy":[96],"directly":[98],"minimizes":[99],"global":[100],"error":[101],"propagation":[102],"across":[103],"entire":[105],"model,":[106],"rather":[107],"than":[108],"isolating":[109],"layer-wise":[110],"errors.":[111],"Building":[112],"this,":[114],"develop":[116],"joint":[119,186],"optimization":[120],"approach":[121],"simultaneously":[123],"learns":[124],"decisions":[127],"policies":[131],"within":[132],"unified":[134],"search":[135],"space.":[136],"Extensive":[137],"experiments":[138],"show":[139],"that,":[140],"at":[141,198],"ultra-low":[142,199],"precisions":[143],"(1-3":[144],"bits),":[145],"our":[146,189],"method":[148,191],"reduces":[149],"WikiText":[150,178],"perplexity":[151,176,194],"by":[152],"up":[153,170],"21%":[155],"compared":[156],"state-of-the-art":[158],"(SoTA)":[159],"weight-activation":[160],"baselines.":[162],"Against":[163],"leading":[164],"weight-only":[165],"methods,":[167],"it":[168],"achieves":[169],"59%":[172],"85%":[174],"lower":[175],"C4,":[180],"respectively.":[181],"Compared":[182],"SoTA":[185],"pruning-and-quantization":[187],"techniques,":[188],"proposed":[190],"delivers":[192],"superior":[193],"reasoning":[196],"performance":[197],"bits.":[200]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-10T00:00:00"}