{"id":"https://openalex.org/W4414196950","doi":"https://doi.org/10.1109/dac63849.2025.11133063","title":"Libra: A Hybrid-Sparse Attention Accelerator Featuring Multi-Level Workload Balance","display_name":"Libra: A Hybrid-Sparse Attention Accelerator Featuring Multi-Level Workload Balance","publication_year":2025,"publication_date":"2025-06-22","ids":{"openalex":"https://openalex.org/W4414196950","doi":"https://doi.org/10.1109/dac63849.2025.11133063"},"language":"en","primary_location":{"id":"doi:10.1109/dac63849.2025.11133063","is_oa":false,"landing_page_url":"https://doi.org/10.1109/dac63849.2025.11133063","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":"article","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":null,"display_name":"Faxian Sun","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":true,"raw_author_name":"Faxian Sun","raw_affiliation_strings":["Fudan University,State Key Lab of Integrated Chips & Systems, and School of Microelectronics,Shanghai,China"],"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/A5111155603","display_name":"Runzhou Zhang","orcid":"https://orcid.org/0009-0004-7843-7752"},"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":"Runzhou Zhang","raw_affiliation_strings":["Fudan University,State Key Lab of Integrated Chips & Systems, and School of Microelectronics,Shanghai,China"],"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/A5100405183","display_name":"Zhenyu Liu","orcid":"https://orcid.org/0000-0003-0251-3267"},"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":"Zhenyu Liu","raw_affiliation_strings":["Fudan University,State Key Lab of Integrated Chips & Systems, and School of Microelectronics,Shanghai,China"],"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/A5068956515","display_name":"Haijun Liao","orcid":"https://orcid.org/0000-0002-5936-9246"},"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":"Heng Liao","raw_affiliation_strings":["Fudan University,State Key Lab of Integrated Chips & Systems, and School of Microelectronics,Shanghai,China"],"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/A5101596001","display_name":"Zhiyong Qin","orcid":"https://orcid.org/0000-0002-7434-7985"},"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":"Zhinan Qin","raw_affiliation_strings":["Fudan University,State Key Lab of Integrated Chips & Systems, and School of Microelectronics,Shanghai,China"],"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"],"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/A5103098516","display_name":"Jun Yu","orcid":"https://orcid.org/0000-0003-4286-9292"},"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"],"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/A5015932760","display_name":"Kun Wang","orcid":"https://orcid.org/0000-0002-7288-1789"},"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"],"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":8,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I4210132426"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.22955733,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"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/T11992","display_name":"CCD and CMOS Imaging Sensors","score":0.9914000034332275,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T11992","display_name":"CCD and CMOS Imaging Sensors","score":0.9914000034332275,"subfield":{"id":"https://openalex.org/subfields/2208","display_name":"Electrical and Electronic Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10904","display_name":"Embedded Systems Design Techniques","score":0.9706000089645386,"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/T10054","display_name":"Parallel Computing and Optimization Techniques","score":0.9550999999046326,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/exploit","display_name":"Exploit","score":0.8169000148773193},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.794700026512146},{"id":"https://openalex.org/keywords/workload","display_name":"Workload","score":0.640999972820282},{"id":"https://openalex.org/keywords/computation","display_name":"Computation","score":0.6381000280380249},{"id":"https://openalex.org/keywords/acceleration","display_name":"Acceleration","score":0.6100999712944031},{"id":"https://openalex.org/keywords/quantization","display_name":"Quantization (signal processing)","score":0.5221999883651733},{"id":"https://openalex.org/keywords/limiting","display_name":"Limiting","score":0.5170999765396118},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5005999803543091}],"concepts":[{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.8169000148773193},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.794700026512146},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7836999893188477},{"id":"https://openalex.org/C2778476105","wikidata":"https://www.wikidata.org/wiki/Q628539","display_name":"Workload","level":2,"score":0.640999972820282},{"id":"https://openalex.org/C45374587","wikidata":"https://www.wikidata.org/wiki/Q12525525","display_name":"Computation","level":2,"score":0.6381000280380249},{"id":"https://openalex.org/C117896860","wikidata":"https://www.wikidata.org/wiki/Q11376","display_name":"Acceleration","level":2,"score":0.6100999712944031},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.5221999883651733},{"id":"https://openalex.org/C188198153","wikidata":"https://www.wikidata.org/wiki/Q1613840","display_name":"Limiting","level":2,"score":0.5170999765396118},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5005999803543091},{"id":"https://openalex.org/C13164978","wikidata":"https://www.wikidata.org/wiki/Q600158","display_name":"Hardware acceleration","level":3,"score":0.44830000400543213},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.38119998574256897},{"id":"https://openalex.org/C2742236","wikidata":"https://www.wikidata.org/wiki/Q924713","display_name":"Efficient energy use","level":2,"score":0.3776000142097473},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.3402999937534332},{"id":"https://openalex.org/C186370098","wikidata":"https://www.wikidata.org/wiki/Q442787","display_name":"Energy (signal processing)","level":2,"score":0.3154999911785126},{"id":"https://openalex.org/C173608175","wikidata":"https://www.wikidata.org/wiki/Q232661","display_name":"Parallel computing","level":1,"score":0.3061000108718872},{"id":"https://openalex.org/C123657996","wikidata":"https://www.wikidata.org/wiki/Q12271","display_name":"Architecture","level":2,"score":0.3057999908924103},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.3034000098705292},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.2822999954223633},{"id":"https://openalex.org/C2780615836","wikidata":"https://www.wikidata.org/wiki/Q2471869","display_name":"USable","level":2,"score":0.2621000111103058},{"id":"https://openalex.org/C179799912","wikidata":"https://www.wikidata.org/wiki/Q205084","display_name":"Computational complexity theory","level":2,"score":0.26179999113082886},{"id":"https://openalex.org/C7366592","wikidata":"https://www.wikidata.org/wiki/Q1255620","display_name":"Dram","level":2,"score":0.25519999861717224},{"id":"https://openalex.org/C149635348","wikidata":"https://www.wikidata.org/wiki/Q193040","display_name":"Embedded system","level":1,"score":0.2542000114917755},{"id":"https://openalex.org/C2780910867","wikidata":"https://www.wikidata.org/wiki/Q1952416","display_name":"Multimodality","level":2,"score":0.25369998812675476}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/dac63849.2025.11133063","is_oa":false,"landing_page_url":"https://doi.org/10.1109/dac63849.2025.11133063","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/F4320321851","display_name":"Fudan University","ror":"https://ror.org/013q1eq08"},{"id":"https://openalex.org/F4320337504","display_name":"Research and Development","ror":"https://ror.org/027s68j25"}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":10,"referenced_works":["https://openalex.org/W2725159389","https://openalex.org/W2949870694","https://openalex.org/W2979826702","https://openalex.org/W3206453033","https://openalex.org/W4280582299","https://openalex.org/W4285236966","https://openalex.org/W4386765191","https://openalex.org/W4387789741","https://openalex.org/W4390097918","https://openalex.org/W4400726540"],"related_works":[],"abstract_inverted_index":{"Transformers":[0],"have":[1,43],"delivered":[2],"exceptional":[3],"performance":[4],"and":[5,27,39,61,126,233],"are":[6],"widely":[7],"used":[8],"across":[9,82],"various":[10],"natural":[11],"language":[12],"processing":[13],"(NLP)":[14],"tasks,":[15],"owing":[16],"to":[17,34,46,54,131,163,218,226],"their":[18],"powerful":[19],"attention":[20,74,105,221],"mechanism.":[21,75],"However,":[22,76],"the":[23,69,73,87,95,112,117,133,151,156,170],"high":[24],"computational":[25],"complexity":[26],"substantial":[28],"memory":[29],"usage":[30],"pose":[31],"significant":[32],"challenges":[33],"inference":[35],"efficiency.":[36,242],"Numerous":[37],"quantization":[38,129],"value-level":[40,60],"sparsification":[41,124,149],"methods":[42],"been":[44],"proposed":[45],"overcome":[47],"these":[48],"challenges.":[49],"Since":[50],"higher":[51],"sparsity":[52,63,78,136,142],"leads":[53],"greater":[55],"acceleration":[56,70,90,96],"efficiency,":[57],"leveraging":[58],"both":[59],"bit-level":[62],"(hybrid":[64],"sparsity)":[65],"can":[66,185],"effectively":[67,168],"exploit":[68,94],"potential":[71,97],"of":[72,89,98,121,135,153],"increased":[77],"exacerbates":[79],"load":[80,171],"imbalance":[81,172],"compute":[83,174],"units,":[84],"potentially":[85],"limiting":[86],"extent":[88],"benefits.":[91],"To":[92,197],"fully":[93],"hybrid":[99,178],"sparsity,":[100],"we":[101,115,159,200],"propose":[102],"Libra,":[103],"an":[104,202],"accelerator":[106],"developed":[107],"through":[108],"algorithm-hardware":[109],"co-design.":[110],"At":[111,155],"algorithm":[113,119],"level,":[114,158],"design":[116,160,201],"bit-group-based":[118],"consisting":[120],"filtered":[122],"bit-group":[123,128],"(FBS)":[125],"dynamic":[127,148],"(DBQ)":[130],"maximize":[132],"utilization":[134],"in":[137,183,188,240],"attention.":[138],"FBS":[139],"imposes":[140],"structured":[141],"on":[143],"weights,":[144],"while":[145],"DBQ":[146,184],"introduces":[147],"during":[150],"computation":[152],"activations.":[154],"hardware":[157],"task":[161],"pool":[162],"achieve":[164],"multi-level":[165],"workload":[166],"balance,":[167],"mitigating":[169],"among":[173],"units":[175],"induced":[176],"by":[177],"sparsity.":[179],"Additionally,":[180],"different":[181],"stages":[182],"be":[186],"executed":[187],"parallel,":[189],"with":[190],"each":[191],"stage":[192],"operating":[193],"at":[194,210],"distinct":[195],"bit-widths.":[196],"support":[198],"this,":[199],"adaptive":[203],"bit-width":[204],"architecture":[205],"that":[206],"enables":[207],"simultaneous":[208],"computations":[209],"varying":[211],"bitwidths.":[212],"Our":[213],"experiments":[214],"demonstrate":[215],"that,":[216],"compared":[217],"state-of-the-art":[219],"(SOTA)":[220],"accelerators,":[222],"Libra":[223],"achieves":[224],"up":[225],"$1.49":[227],"\\times":[228,235],"\\sim":[229,236],"5.89":[230],"\\times$":[231,238],"speedup":[232],"$2.65":[234],"10.82":[237],"enhancement":[239],"energy":[241]},"counts_by_year":[],"updated_date":"2026-04-17T18:11:37.981687","created_date":"2025-10-10T00:00:00"}