{"id":"https://openalex.org/W7165648526","doi":"https://doi.org/10.48550/arxiv.2606.21257","title":"An Empirical Study of OpenPangu Quantization on Ascend NPUs","display_name":"An Empirical Study of OpenPangu Quantization on Ascend NPUs","publication_year":2026,"publication_date":"2026-06-19","ids":{"openalex":"https://openalex.org/W7165648526","doi":"https://doi.org/10.48550/arxiv.2606.21257"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.21257","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.21257","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.2606.21257","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5100519813","display_name":"Tong Shi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shi, Tong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139167792","display_name":"Jiacheng Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Jiacheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139140248","display_name":"Hui Xie","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xie, Hui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139136042","display_name":"Ying Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Ying","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139168957","display_name":"Aishan Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Aishan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5139216555","display_name":"Jinyang Guo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guo, Jinyang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5139142164","display_name":"Xianglong Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Xianglong","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/T10028","display_name":"Topic Modeling","score":0.19609999656677246,"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/T10028","display_name":"Topic Modeling","score":0.19609999656677246,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.17069999873638153,"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":0.07519999891519547,"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":0.7971000075340271},{"id":"https://openalex.org/keywords/perplexity","display_name":"Perplexity","score":0.6693999767303467},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.5281999707221985},{"id":"https://openalex.org/keywords/empirical-research","display_name":"Empirical research","score":0.5227000117301941},{"id":"https://openalex.org/keywords/binary-number","display_name":"Binary number","score":0.5069000124931335}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.7971000075340271},{"id":"https://openalex.org/C100279451","wikidata":"https://www.wikidata.org/wiki/Q372193","display_name":"Perplexity","level":3,"score":0.6693999767303467},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.635200023651123},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.5281999707221985},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.5227000117301941},{"id":"https://openalex.org/C48372109","wikidata":"https://www.wikidata.org/wiki/Q3913","display_name":"Binary number","level":2,"score":0.5069000124931335},{"id":"https://openalex.org/C81081738","wikidata":"https://www.wikidata.org/wiki/Q55542","display_name":"Lossless compression","level":3,"score":0.45329999923706055},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.38940000534057617},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3384000062942505},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.2782000005245209},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.2689000070095062},{"id":"https://openalex.org/C43126263","wikidata":"https://www.wikidata.org/wiki/Q128751","display_name":"Source code","level":2,"score":0.2554999887943268},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2533999979496002}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.21257","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.21257","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.2606.21257","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.21257","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":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"OpenPangu":[0,34,140],"models":[1,38],"are":[2],"attractive":[3],"targets":[4],"for":[5,82,90,99,138],"private":[6],"and":[7,36,48,60,66,106,115,122,143],"domestic":[8],"large-language-model":[9],"deployment,":[10],"yet":[11],"their":[12],"robustness":[13],"under":[14,62],"aggressive":[15],"post-training":[16,50],"quantization":[17,51,78,87,141],"on":[18,39,103],"Ascend":[19,41],"NPUs":[20],"has":[21],"not":[22],"been":[23],"systematically":[24],"characterized.":[25],"This":[26],"paper":[27],"conducts":[28],"a":[29,63],"controlled":[30],"empirical":[31],"study":[32],"of":[33,148],"1B":[35,101],"7B":[37,92],"Huawei":[40],"910B1":[42],"NPUs.":[43],"We":[44],"evaluate":[45],"representative":[46],"weight-only":[47,77],"weight-activation":[49],"methods,":[52],"including":[53],"RTN,":[54],"GPTQ,":[55],"AWQ,":[56],"SmoothQuant,":[57],"GPTAQ,":[58],"BiLLM,":[59],"SliM-LLM,":[61],"unified":[64],"calibration":[65],"evaluation":[67,71],"protocol.":[68],"Across":[69],"18":[70],"tasks,":[72],"we":[73],"find":[74],"that":[75],"8-bit":[76],"is":[79,95],"effectively":[80],"lossless":[81],"both":[83],"models,":[84],"while":[85],"4-bit":[86],"remains":[88,111],"practical":[89],"the":[91,100,145],"model":[93,102],"but":[94],"visibly":[96],"more":[97],"harmful":[98],"reasoning,":[104],"math,":[105],"code":[107],"tasks.":[108],"Ultra-low":[109],"precision":[110],"challenging:":[112],"most":[113],"2-bit":[114],"binary":[116],"settings":[117,142],"collapse":[118],"to":[119],"near-random":[120],"behavior,":[121],"W4A4":[123],"SmoothQuant":[124],"produces":[125],"non-finite":[126],"perplexity":[127],"in":[128],"our":[129],"evaluation.":[130],"These":[131],"results":[132],"provide":[133],"an":[134],"NPU-oriented":[135],"accuracy":[136],"map":[137],"selecting":[139],"highlight":[144],"persistent":[146],"difficulty":[147],"extreme":[149],"low-bit":[150],"compression.":[151]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-24T00:00:00"}