{"id":"https://openalex.org/W4414197938","doi":"https://doi.org/10.1109/dac63849.2025.11133363","title":"XShift: FPGA-efficient Binarized LLM with Joint Quantization and Sparsification","display_name":"XShift: FPGA-efficient Binarized LLM with Joint Quantization and Sparsification","publication_year":2025,"publication_date":"2025-06-22","ids":{"openalex":"https://openalex.org/W4414197938","doi":"https://doi.org/10.1109/dac63849.2025.11133363"},"language":"en","primary_location":{"id":"doi:10.1109/dac63849.2025.11133363","is_oa":false,"landing_page_url":"https://doi.org/10.1109/dac63849.2025.11133363","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 62nd ACM/IEEE Design Automation Conference (DAC)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101954794","display_name":"Shuai Zhou","orcid":"https://orcid.org/0009-0007-1531-3447"},"institutions":[{"id":"https://openalex.org/I4210132426","display_name":"Shanghai Fudan Microelectronics (China)","ror":"https://ror.org/02vfj3j86","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210132426"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Shuai Zhou","raw_affiliation_strings":["Fudan University,State Key Lab of Integrated Chips & Systems, and School of Microelectronics,Shanghai,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University,State Key Lab of Integrated Chips & Systems, and School of Microelectronics,Shanghai,China","institution_ids":["https://openalex.org/I4210132426"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5022672817","display_name":"Huinan Tian","orcid":"https://orcid.org/0009-0002-3887-6852"},"institutions":[{"id":"https://openalex.org/I4210132426","display_name":"Shanghai Fudan Microelectronics (China)","ror":"https://ror.org/02vfj3j86","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210132426"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Huinan Tian","raw_affiliation_strings":["Fudan University,State Key Lab of Integrated Chips & Systems, and School of Microelectronics,Shanghai,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University,State Key Lab of Integrated Chips & Systems, and School of Microelectronics,Shanghai,China","institution_ids":["https://openalex.org/I4210132426"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5117084707","display_name":"Sisi Meng","orcid":null},"institutions":[{"id":"https://openalex.org/I4210132426","display_name":"Shanghai Fudan Microelectronics (China)","ror":"https://ror.org/02vfj3j86","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210132426"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Sisi Meng","raw_affiliation_strings":["Fudan University,State Key Lab of Integrated Chips & Systems, and School of Microelectronics,Shanghai,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University,State Key Lab of Integrated Chips & Systems, and School of Microelectronics,Shanghai,China","institution_ids":["https://openalex.org/I4210132426"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101854580","display_name":"Jianli Chen","orcid":"https://orcid.org/0000-0002-1391-2696"},"institutions":[{"id":"https://openalex.org/I4210132426","display_name":"Shanghai Fudan Microelectronics (China)","ror":"https://ror.org/02vfj3j86","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210132426"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jianli Chen","raw_affiliation_strings":["Fudan University,State Key Lab of Integrated Chips & Systems, and School of Microelectronics,Shanghai,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University,State Key Lab of Integrated Chips & Systems, and School of Microelectronics,Shanghai,China","institution_ids":["https://openalex.org/I4210132426"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100614702","display_name":"Jun Yu","orcid":"https://orcid.org/0000-0001-5008-2153"},"institutions":[{"id":"https://openalex.org/I4210132426","display_name":"Shanghai Fudan Microelectronics (China)","ror":"https://ror.org/02vfj3j86","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210132426"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jun Yu","raw_affiliation_strings":["Fudan University,State Key Lab of Integrated Chips & Systems, and School of Microelectronics,Shanghai,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University,State Key Lab of Integrated Chips & Systems, and School of Microelectronics,Shanghai,China","institution_ids":["https://openalex.org/I4210132426"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5100366692","display_name":"Kun Wang","orcid":"https://orcid.org/0000-0002-9099-2781"},"institutions":[{"id":"https://openalex.org/I4210132426","display_name":"Shanghai Fudan Microelectronics (China)","ror":"https://ror.org/02vfj3j86","country_code":"CN","type":"company","lineage":["https://openalex.org/I4210132426"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Kun Wang","raw_affiliation_strings":["Fudan University,State Key Lab of Integrated Chips & Systems, and School of Microelectronics,Shanghai,China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Fudan University,State Key Lab of Integrated Chips & Systems, and School of Microelectronics,Shanghai,China","institution_ids":["https://openalex.org/I4210132426"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I4210132426"],"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":"1","last_page":"7"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11269","display_name":"Algorithms and Data Compression","score":0.9294000267982483,"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/T11269","display_name":"Algorithms and Data Compression","score":0.9294000267982483,"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/T10601","display_name":"Handwritten Text Recognition Techniques","score":0.9228000044822693,"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/xnor-gate","display_name":"XNOR gate","score":0.6646000146865845},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.6366000175476074},{"id":"https://openalex.org/keywords/field-programmable-gate-array","display_name":"Field-programmable gate array","score":0.6122999787330627},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.6047999858856201},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.5983999967575073},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5091000199317932},{"id":"https://openalex.org/keywords/computational-complexity-theory","display_name":"Computational complexity theory","score":0.4397999942302704},{"id":"https://openalex.org/keywords/exploit","display_name":"Exploit","score":0.43320000171661377},{"id":"https://openalex.org/keywords/reduction","display_name":"Reduction (mathematics)","score":0.4207000136375427},{"id":"https://openalex.org/keywords/softmax-function","display_name":"Softmax function","score":0.4052000045776367}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7554000020027161},{"id":"https://openalex.org/C57684291","wikidata":"https://www.wikidata.org/wiki/Q1336142","display_name":"XNOR gate","level":4,"score":0.6646000146865845},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.6366000175476074},{"id":"https://openalex.org/C42935608","wikidata":"https://www.wikidata.org/wiki/Q190411","display_name":"Field-programmable gate array","level":2,"score":0.6122999787330627},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.6047999858856201},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.5983999967575073},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5091000199317932},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.46050000190734863},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.4397999942302704},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.43320000171661377},{"id":"https://openalex.org/C111335779","wikidata":"https://www.wikidata.org/wiki/Q3454686","display_name":"Reduction (mathematics)","level":2,"score":0.4207000136375427},{"id":"https://openalex.org/C188441871","wikidata":"https://www.wikidata.org/wiki/Q7554146","display_name":"Softmax function","level":3,"score":0.4052000045776367},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.37940001487731934},{"id":"https://openalex.org/C57273362","wikidata":"https://www.wikidata.org/wiki/Q576722","display_name":"Decoding methods","level":2,"score":0.37709999084472656},{"id":"https://openalex.org/C125411270","wikidata":"https://www.wikidata.org/wiki/Q18653","display_name":"Encoding (memory)","level":2,"score":0.37610000371932983},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.37599998712539673},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.37279999256134033},{"id":"https://openalex.org/C150741067","wikidata":"https://www.wikidata.org/wiki/Q2377218","display_name":"Systolic array","level":3,"score":0.3716999888420105},{"id":"https://openalex.org/C13164978","wikidata":"https://www.wikidata.org/wiki/Q600158","display_name":"Hardware acceleration","level":3,"score":0.3686000108718872},{"id":"https://openalex.org/C84462506","wikidata":"https://www.wikidata.org/wiki/Q173142","display_name":"Digital signal processing","level":2,"score":0.3637999892234802},{"id":"https://openalex.org/C2778584072","wikidata":"https://www.wikidata.org/wiki/Q7353545","display_name":"Robustification","level":3,"score":0.34619998931884766},{"id":"https://openalex.org/C2221639","wikidata":"https://www.wikidata.org/wiki/Q2877","display_name":"Discrete cosine transform","level":3,"score":0.3407999873161316},{"id":"https://openalex.org/C9390403","wikidata":"https://www.wikidata.org/wiki/Q3966","display_name":"Computer hardware","level":1,"score":0.34060001373291016},{"id":"https://openalex.org/C2780971903","wikidata":"https://www.wikidata.org/wiki/Q2933705","display_name":"Stochastic computing","level":3,"score":0.33500000834465027},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.3314000070095062},{"id":"https://openalex.org/C162307627","wikidata":"https://www.wikidata.org/wiki/Q204833","display_name":"Enhanced Data Rates for GSM Evolution","level":2,"score":0.33009999990463257},{"id":"https://openalex.org/C81081738","wikidata":"https://www.wikidata.org/wiki/Q55542","display_name":"Lossless compression","level":3,"score":0.32910001277923584},{"id":"https://openalex.org/C79337645","wikidata":"https://www.wikidata.org/wiki/Q779824","display_name":"Outlier","level":2,"score":0.328900009393692},{"id":"https://openalex.org/C79974875","wikidata":"https://www.wikidata.org/wiki/Q483639","display_name":"Cloud computing","level":2,"score":0.3264999985694885},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.3206000030040741},{"id":"https://openalex.org/C2778456923","wikidata":"https://www.wikidata.org/wiki/Q5337692","display_name":"Edge computing","level":3,"score":0.31940001249313354},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.3068000078201294},{"id":"https://openalex.org/C127964446","wikidata":"https://www.wikidata.org/wiki/Q1092142","display_name":"Computational resource","level":3,"score":0.30570000410079956},{"id":"https://openalex.org/C179518139","wikidata":"https://www.wikidata.org/wiki/Q5140297","display_name":"Coding (social sciences)","level":2,"score":0.28929999470710754},{"id":"https://openalex.org/C459310","wikidata":"https://www.wikidata.org/wiki/Q117801","display_name":"Computational science","level":1,"score":0.28859999775886536},{"id":"https://openalex.org/C163294075","wikidata":"https://www.wikidata.org/wiki/Q581861","display_name":"Noise reduction","level":2,"score":0.2782000005245209},{"id":"https://openalex.org/C519991488","wikidata":"https://www.wikidata.org/wiki/Q28865","display_name":"Python (programming language)","level":2,"score":0.27730000019073486},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.27140000462532043},{"id":"https://openalex.org/C2778770139","wikidata":"https://www.wikidata.org/wiki/Q1966904","display_name":"Solver","level":2,"score":0.266400009393692},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.26579999923706055}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/dac63849.2025.11133363","is_oa":false,"landing_page_url":"https://doi.org/10.1109/dac63849.2025.11133363","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 62nd ACM/IEEE Design Automation Conference (DAC)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320337504","display_name":"Research and Development","ror":"https://ror.org/027s68j25"}],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":20,"referenced_works":["https://openalex.org/W2883920103","https://openalex.org/W2966983573","https://openalex.org/W3100985894","https://openalex.org/W3130641740","https://openalex.org/W4293025109","https://openalex.org/W4308083513","https://openalex.org/W4366341968","https://openalex.org/W4380881077","https://openalex.org/W4385815501","https://openalex.org/W4389519530","https://openalex.org/W4392181806","https://openalex.org/W4393147854","https://openalex.org/W4393578753","https://openalex.org/W4401211704","https://openalex.org/W4401211807","https://openalex.org/W4404133534","https://openalex.org/W4404133703","https://openalex.org/W4404134154","https://openalex.org/W4409285558","https://openalex.org/W4409362928"],"related_works":[],"abstract_inverted_index":{"Binarization":[0],"is":[1],"a":[2,62,89,132,172,203],"promising":[3],"approach":[4],"to":[5,25,33,37,150,214],"significantly":[6],"reduce":[7],"computational":[8,51,154],"complexity":[9],"by":[10],"replacing":[11],"multiplications":[12,98],"with":[13],"hardwareefficient":[14],"XNOR":[15,100],"operations.":[16],"However,":[17],"the":[18,38,56,152,162,222],"binarization":[19,31],"of":[20,156,199,224,229],"LLM":[21,79,217],"activations":[22],"often":[23],"leads":[24],"severe":[26],"accuracy":[27],"degradation,":[28],"while":[29,184],"weight-only":[30],"fails":[32],"eliminate":[34],"multipliers":[35],"due":[36],"Self-Attention":[39],"mechanism.":[40],"Furthermore,":[41],"LLMs":[42],"exhibit":[43],"distinctive":[44],"channel-level":[45,118],"data":[46,119],"distribution":[47,120],"characteristics":[48],"and":[49,52,58,101,106,112,121,128,130,142,165,188,202],"differing":[50],"memory":[53],"requirements":[54],"between":[55],"Pre-fill":[57],"Decoding":[59],"stages,":[60],"necessitating":[61],"specialized":[63],"inference":[64,80,191],"framework.":[65],"In":[66],"response,":[67],"we":[68],"introduce":[69],"XShift,":[70],"an":[71,137,143,196],"algorithm-hardware":[72],"co-design":[73],"framework":[74],"optimized":[75,126],"for":[76,125,226],"efficient":[77],"binarized":[78,157],"on":[81,161],"FPGAs.":[82],"XShift":[83,170,194,225],"incorporates":[84],"three":[85],"key":[86],"contributions:":[87],"(1)":[88],"hardwarefriendly":[90],"XNOR-Shift":[91,138],"Encoding":[92],"(XSE)":[93],"format":[94],"that":[95,169],"transforms":[96],"traditional":[97],"into":[99],"shift":[102],"operations,":[103],"ensuring":[104],"scalability":[105],"precision;":[107],"(2)":[108],"Hardware":[109],"Adaptive":[110],"Outlier":[111],"Sparsity":[113],"(HAOS)":[114],"techniques,":[115],"which":[116],"exploit":[117],"systolic":[122],"array":[123],"architectures":[124],"quantization":[127],"sparsification;":[129],"(3)":[131],"dedicated":[133],"hardware":[134],"accelerator":[135],"featuring":[136],"Systolic":[139],"Array":[140],"(XSSA)":[141],"enhanced":[144],"Base-2":[145],"SoftMax":[146],"Converter":[147],"(BSMC),":[148],"designed":[149],"address":[151],"specific":[153],"demands":[155],"LLMs.":[158,230],"Experimental":[159],"evaluations":[160],"Alveo":[163],"U280":[164],"U50":[166],"FPGA":[167],"demonstrate":[168],"achieves":[171],"$\\mathbf{1":[173],"0":[174],"-":[175],"1":[176],"5":[177],"x}$":[178],"reduction":[179],"in":[180,190,207],"DSP":[181],"resource":[182],"usage":[183],"surpassing":[185],"existing":[186],"accelerators":[187],"GPUs":[189],"performance.":[192],"Specifically,":[193],"delivers":[195],"average":[197],"speedup":[198],"4.174.76":[200],"x":[201,205],"14.29-6.95":[204],"improvement":[206],"energy":[208],"efficiency,":[209],"alongside":[210],"lower":[211],"perplexity":[212],"compared":[213],"other":[215],"low-precision":[216],"techniques.":[218],"These":[219],"results":[220],"underscore":[221],"potential":[223],"edge":[227],"deployment":[228]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}