{"id":"https://openalex.org/W7134907250","doi":"https://doi.org/10.1145/3779212.3790185","title":"M <sup>2</sup> XFP: A Metadata-Augmented Microscaling Data Format for Efficient Low-bit Quantization","display_name":"M <sup>2</sup> XFP: A Metadata-Augmented Microscaling Data Format for Efficient Low-bit Quantization","publication_year":2026,"publication_date":"2026-03-10","ids":{"openalex":"https://openalex.org/W7134907250","doi":"https://doi.org/10.1145/3779212.3790185"},"language":null,"primary_location":{"id":"doi:10.1145/3779212.3790185","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3779212.3790185","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 31st ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 2","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://doi.org/10.1145/3779212.3790185","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101721549","display_name":"Weiming Hu","orcid":"https://orcid.org/0009-0003-5115-0498"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Weiming Hu","raw_affiliation_strings":["Shanghai Jiao Tong University, Shanghai, China and Shanghai Qi Zhi Institute, Shanghai, China"],"raw_orcid":"https://orcid.org/0009-0003-5115-0498","affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University, Shanghai, China and Shanghai Qi Zhi Institute, Shanghai, China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128744161","display_name":"Zihan Zhang","orcid":null},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zihan Zhang","raw_affiliation_strings":["Shanghai Jiao Tong University, Shanghai, China"],"raw_orcid":"https://orcid.org/0009-0008-7683-2934","affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University, Shanghai, China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5040260713","display_name":"H. Zhang","orcid":null},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Haoyan Zhang","raw_affiliation_strings":["Shanghai Jiao Tong University, Shanghai, China and Shanghai Qi Zhi Institute, Shanghai, China"],"raw_orcid":"https://orcid.org/0009-0009-8634-5395","affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University, Shanghai, China and Shanghai Qi Zhi Institute, Shanghai, China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128769521","display_name":"Chen Zhang","orcid":null},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Chen Zhang","raw_affiliation_strings":["Shanghai Jiao Tong University, Shanghai, China"],"raw_orcid":"https://orcid.org/0000-0003-2762-2726","affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University, Shanghai, China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128735393","display_name":"Cong Guo","orcid":null},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Cong Guo","raw_affiliation_strings":["Shanghai Jiao Tong University, Shanghai, China"],"raw_orcid":"https://orcid.org/0000-0002-4479-5525","affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University, Shanghai, China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101540583","display_name":"Yu Feng","orcid":"https://orcid.org/0000-0002-2192-5737"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Yu Feng","raw_affiliation_strings":["Shanghai Jiao Tong University, Shanghai, China"],"raw_orcid":"https://orcid.org/0000-0002-2192-5737","affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University, Shanghai, China","institution_ids":["https://openalex.org/I183067930"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5111768283","display_name":"Tianchi Hu","orcid":null},"institutions":[{"id":"https://openalex.org/I2250955327","display_name":"Huawei Technologies (China)","ror":"https://ror.org/00cmhce21","country_code":"CN","type":"company","lineage":["https://openalex.org/I2250955327"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Tianchi Hu","raw_affiliation_strings":["Computing Product Line, Huawei, Shanghai, China"],"raw_orcid":"https://orcid.org/0009-0004-2986-9858","affiliations":[{"raw_affiliation_string":"Computing Product Line, Huawei, Shanghai, China","institution_ids":["https://openalex.org/I2250955327"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124081211","display_name":"Guanglin Li","orcid":null},"institutions":[{"id":"https://openalex.org/I2250955327","display_name":"Huawei Technologies (China)","ror":"https://ror.org/00cmhce21","country_code":"CN","type":"company","lineage":["https://openalex.org/I2250955327"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Guanglin Li","raw_affiliation_strings":["Computing Product Line, Huawei, Shanghai, China"],"raw_orcid":"https://orcid.org/0009-0000-8996-3775","affiliations":[{"raw_affiliation_string":"Computing Product Line, Huawei, Shanghai, China","institution_ids":["https://openalex.org/I2250955327"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124086334","display_name":"Guipeng Hu","orcid":null},"institutions":[{"id":"https://openalex.org/I2250955327","display_name":"Huawei Technologies (China)","ror":"https://ror.org/00cmhce21","country_code":"CN","type":"company","lineage":["https://openalex.org/I2250955327"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Guipeng Hu","raw_affiliation_strings":["Computing Product Line, Huawei, Shanghai, China"],"raw_orcid":"https://orcid.org/0009-0007-7721-0048","affiliations":[{"raw_affiliation_string":"Computing Product Line, Huawei, Shanghai, China","institution_ids":["https://openalex.org/I2250955327"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124061494","display_name":"Junsong Wang","orcid":null},"institutions":[{"id":"https://openalex.org/I2250955327","display_name":"Huawei Technologies (China)","ror":"https://ror.org/00cmhce21","country_code":"CN","type":"company","lineage":["https://openalex.org/I2250955327"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Junsong Wang","raw_affiliation_strings":["Computing Product Line, Huawei, Beijing, China"],"raw_orcid":"https://orcid.org/0009-0006-5954-6958","affiliations":[{"raw_affiliation_string":"Computing Product Line, Huawei, Beijing, China","institution_ids":["https://openalex.org/I2250955327"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5003939279","display_name":"Jingwen Leng","orcid":"https://orcid.org/0000-0002-5660-5493"},"institutions":[{"id":"https://openalex.org/I183067930","display_name":"Shanghai Jiao Tong University","ror":"https://ror.org/0220qvk04","country_code":"CN","type":"education","lineage":["https://openalex.org/I183067930"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jingwen Leng","raw_affiliation_strings":["Shanghai Jiao Tong University, Shanghai, China and Shanghai Qi Zhi Institute, Shanghai, China"],"raw_orcid":"https://orcid.org/0000-0002-5660-5493","affiliations":[{"raw_affiliation_string":"Shanghai Jiao Tong University, Shanghai, China and Shanghai Qi Zhi Institute, Shanghai, China","institution_ids":["https://openalex.org/I183067930"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":11,"corresponding_author_ids":["https://openalex.org/A5101721549"],"corresponding_institution_ids":["https://openalex.org/I183067930"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.41290805,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"1151","last_page":"1167"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.4934000074863434,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.4934000074863434,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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/T10904","display_name":"Embedded Systems Design Techniques","score":0.13109999895095825,"subfield":{"id":"https://openalex.org/subfields/1708","display_name":"Hardware and Architecture"},"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":0.08630000054836273,"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":0.8148000240325928},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.5874999761581421},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.492000013589859},{"id":"https://openalex.org/keywords/metadata","display_name":"Metadata","score":0.45730000734329224},{"id":"https://openalex.org/keywords/reduction","display_name":"Reduction (mathematics)","score":0.4514999985694885},{"id":"https://openalex.org/keywords/data-reduction","display_name":"Data reduction","score":0.35190001130104065},{"id":"https://openalex.org/keywords/efficient-energy-use","display_name":"Efficient energy use","score":0.3303999900817871}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.8148000240325928},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.758899986743927},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.5874999761581421},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.5479999780654907},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.492000013589859},{"id":"https://openalex.org/C93518851","wikidata":"https://www.wikidata.org/wiki/Q180160","display_name":"Metadata","level":2,"score":0.45730000734329224},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.4514999985694885},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.40369999408721924},{"id":"https://openalex.org/C153914771","wikidata":"https://www.wikidata.org/wiki/Q5227343","display_name":"Data reduction","level":2,"score":0.35190001130104065},{"id":"https://openalex.org/C2742236","wikidata":"https://www.wikidata.org/wiki/Q924713","display_name":"Efficient energy use","level":2,"score":0.3303999900817871},{"id":"https://openalex.org/C2780586882","wikidata":"https://www.wikidata.org/wiki/Q7520643","display_name":"Simple (philosophy)","level":2,"score":0.3264000117778778},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.30790001153945923},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.3003999888896942},{"id":"https://openalex.org/C186370098","wikidata":"https://www.wikidata.org/wiki/Q442787","display_name":"Energy (signal processing)","level":2,"score":0.28690001368522644},{"id":"https://openalex.org/C2781039887","wikidata":"https://www.wikidata.org/wiki/Q1391724","display_name":"Factor (programming language)","level":2,"score":0.2849999964237213},{"id":"https://openalex.org/C2780056265","wikidata":"https://www.wikidata.org/wiki/Q106239881","display_name":"High dynamic range","level":3,"score":0.2563999891281128},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.2554999887943268},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.2535000145435333},{"id":"https://openalex.org/C94375191","wikidata":"https://www.wikidata.org/wiki/Q11205","display_name":"Arithmetic","level":1,"score":0.2531000077724457}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1145/3779212.3790185","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3779212.3790185","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 31st ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 2","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.1145/3779212.3790185","is_oa":true,"landing_page_url":"https://doi.org/10.1145/3779212.3790185","pdf_url":null,"source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Proceedings of the 31st ACM International Conference on Architectural Support for Programming Languages and Operating Systems, Volume 2","raw_type":"proceedings-article"},"sustainable_development_goals":[{"display_name":"Affordable and clean energy","score":0.8929111957550049,"id":"https://metadata.un.org/sdg/7"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":27,"referenced_works":["https://openalex.org/W2562773490","https://openalex.org/W2725159389","https://openalex.org/W2946609015","https://openalex.org/W2946659172","https://openalex.org/W2998617917","https://openalex.org/W3194676777","https://openalex.org/W4308083739","https://openalex.org/W4366341968","https://openalex.org/W4380874652","https://openalex.org/W4387321091","https://openalex.org/W4389491858","https://openalex.org/W4394998892","https://openalex.org/W4402475796","https://openalex.org/W4402475821","https://openalex.org/W4404133534","https://openalex.org/W4407217670","https://openalex.org/W4409248487","https://openalex.org/W4409248488","https://openalex.org/W4409248600","https://openalex.org/W4409248709","https://openalex.org/W4411486471","https://openalex.org/W4411486493","https://openalex.org/W4411688651","https://openalex.org/W4414197630","https://openalex.org/W4415308961","https://openalex.org/W6893007900","https://openalex.org/W7133233472"],"related_works":[],"abstract_inverted_index":{"Existing":[0],"low-bit":[1],"Microscaling":[2],"(MX)":[3],"formats,":[4],"such":[5],"as":[6],"MXFP4,":[7],"often":[8],"suffer":[9],"from":[10],"substantial":[11],"accuracy":[12,39,100,109],"degradation":[13],"due":[14],"to":[15,37,112,119,131],"the":[16,24,75,89,99,120],"use":[17],"of":[18,52],"a":[19,49,58,81,105,115],"shared":[20],"scaling":[21],"factor":[22],"with":[23,70],"Power-of-Two":[25],"format.":[26],"In":[27],"this":[28],"work,":[29],"we":[30,79],"explore":[31],"strategies":[32],"that":[33,94],"introduce":[34],"minimal":[35],"metadata":[36],"recover":[38],"lost":[40],"during":[41],"quantization":[42,69],"while":[43],"maintaining":[44],"high":[45],"bit":[46],"efficiency":[47],"across":[48],"wide":[50],"range":[51],"large":[53],"language":[54],"models.":[55],"We":[56],"propose":[57],"complete":[59],"algorithm-hardware":[60],"co-design":[61],"based":[62],"on":[63,103,123],"flexible":[64],"metadata,":[65],"featuring":[66],"an":[67],"online":[68],"simple":[71],"encoding.":[72],"To":[73],"support":[74],"proposed":[76],"method":[77,96],"efficiently,":[78],"implement":[80],"lightweight":[82],"hardware":[83],"unit":[84],"and":[85,114,134],"integrate":[86],"it":[87],"into":[88],"accelerator.":[90],"Evaluation":[91],"results":[92],"demonstrate":[93],"our":[95,127],"substantially":[97],"narrows":[98],"gap,":[101],"achieving":[102],"average":[104],"70.63%":[106],"reduction":[107,117],"in":[108],"loss":[110],"compared":[111],"MXFP4":[113],"37.30%":[116],"relative":[118],"latest":[121],"NVFP4":[122],"LLM":[124],"benchmarks.":[125],"Furthermore,":[126],"design":[128],"delivers":[129],"up":[130],"1.91\u00d7":[132],"speedup":[133],"1.75\u00d7":[135],"energy":[136],"savings":[137],"over":[138],"state-of-the-art":[139],"accelerators.":[140]},"counts_by_year":[],"updated_date":"2026-03-12T06:18:43.230356","created_date":"2026-03-12T00:00:00"}