{"id":"https://openalex.org/W4416429839","doi":"https://doi.org/10.1109/iccad66269.2025.11240685","title":"Squat: Quant Small Language Models on the Edge","display_name":"Squat: Quant Small Language Models on the Edge","publication_year":2025,"publication_date":"2025-10-26","ids":{"openalex":"https://openalex.org/W4416429839","doi":"https://doi.org/10.1109/iccad66269.2025.11240685"},"language":null,"primary_location":{"id":"doi:10.1109/iccad66269.2025.11240685","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iccad66269.2025.11240685","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/ACM International Conference On Computer Aided Design (ICCAD)","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":"https://openalex.org/A5101655527","display_name":"Xuan Shen","orcid":"https://orcid.org/0000-0003-4965-7321"},"institutions":[{"id":"https://openalex.org/I87182695","display_name":"Universidad del Noreste","ror":"https://ror.org/02ahky613","country_code":"MX","type":"education","lineage":["https://openalex.org/I87182695"]}],"countries":["MX"],"is_corresponding":false,"raw_author_name":"Xuan Shen","raw_affiliation_strings":["Northeastern University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Northeastern University","institution_ids":["https://openalex.org/I87182695"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5089349472","display_name":"Peiyan Dong","orcid":"https://orcid.org/0000-0001-5287-5149"},"institutions":[{"id":"https://openalex.org/I4210109586","display_name":"Moscow Institute of Thermal Technology","ror":"https://ror.org/021es5e59","country_code":"RU","type":"facility","lineage":["https://openalex.org/I4210109586"]}],"countries":["RU"],"is_corresponding":false,"raw_author_name":"Peiyan Dong","raw_affiliation_strings":["MIT"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"MIT","institution_ids":["https://openalex.org/I4210109586"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5078971265","display_name":"Zhenglun Kong","orcid":"https://orcid.org/0000-0002-8120-4456"},"institutions":[{"id":"https://openalex.org/I2801851002","display_name":"Harvard University Press","ror":"https://ror.org/006v7bf86","country_code":"US","type":"other","lineage":["https://openalex.org/I136199984","https://openalex.org/I2801851002"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Zhenglun Kong","raw_affiliation_strings":["Harvard University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Harvard University","institution_ids":["https://openalex.org/I2801851002"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5101928537","display_name":"Yifan Gong","orcid":"https://orcid.org/0000-0002-3912-097X"},"institutions":[{"id":"https://openalex.org/I87182695","display_name":"Universidad del Noreste","ror":"https://ror.org/02ahky613","country_code":"MX","type":"education","lineage":["https://openalex.org/I87182695"]}],"countries":["MX"],"is_corresponding":false,"raw_author_name":"Yifan Gong","raw_affiliation_strings":["Northeastern University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Northeastern University","institution_ids":["https://openalex.org/I87182695"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5033198334","display_name":"Changdi Yang","orcid":"https://orcid.org/0000-0002-8848-3806"},"institutions":[{"id":"https://openalex.org/I87182695","display_name":"Universidad del Noreste","ror":"https://ror.org/02ahky613","country_code":"MX","type":"education","lineage":["https://openalex.org/I87182695"]}],"countries":["MX"],"is_corresponding":false,"raw_author_name":"Changdi Yang","raw_affiliation_strings":["Northeastern University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Northeastern University","institution_ids":["https://openalex.org/I87182695"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5067112514","display_name":"Zhaoyang Han","orcid":"https://orcid.org/0000-0002-6816-6555"},"institutions":[{"id":"https://openalex.org/I87182695","display_name":"Universidad del Noreste","ror":"https://ror.org/02ahky613","country_code":"MX","type":"education","lineage":["https://openalex.org/I87182695"]}],"countries":["MX"],"is_corresponding":false,"raw_author_name":"Zhaoyang Han","raw_affiliation_strings":["Northeastern University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Northeastern University","institution_ids":["https://openalex.org/I87182695"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5088784101","display_name":"Yanyue Xie","orcid":"https://orcid.org/0000-0002-4325-521X"},"institutions":[{"id":"https://openalex.org/I87182695","display_name":"Universidad del Noreste","ror":"https://ror.org/02ahky613","country_code":"MX","type":"education","lineage":["https://openalex.org/I87182695"]}],"countries":["MX"],"is_corresponding":false,"raw_author_name":"Yanyue Xie","raw_affiliation_strings":["Northeastern University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Northeastern University","institution_ids":["https://openalex.org/I87182695"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100858653","display_name":"Lei L\u00fc","orcid":null},"institutions":[{"id":"https://openalex.org/I87182695","display_name":"Universidad del Noreste","ror":"https://ror.org/02ahky613","country_code":"MX","type":"education","lineage":["https://openalex.org/I87182695"]}],"countries":["MX"],"is_corresponding":false,"raw_author_name":"Lei Lu","raw_affiliation_strings":["Northeastern University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Northeastern University","institution_ids":["https://openalex.org/I87182695"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100303246","display_name":"Cheng Lyu","orcid":"https://orcid.org/0000-0003-4270-0702"},"institutions":[{"id":"https://openalex.org/I87182695","display_name":"Universidad del Noreste","ror":"https://ror.org/02ahky613","country_code":"MX","type":"education","lineage":["https://openalex.org/I87182695"]}],"countries":["MX"],"is_corresponding":false,"raw_author_name":"Cheng Lyu","raw_affiliation_strings":["Northeastern University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Northeastern University","institution_ids":["https://openalex.org/I87182695"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5038380400","display_name":"Chao Wu","orcid":"https://orcid.org/0000-0003-1551-9676"},"institutions":[{"id":"https://openalex.org/I87182695","display_name":"Universidad del Noreste","ror":"https://ror.org/02ahky613","country_code":"MX","type":"education","lineage":["https://openalex.org/I87182695"]}],"countries":["MX"],"is_corresponding":false,"raw_author_name":"Chao Wu","raw_affiliation_strings":["Northeastern University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Northeastern University","institution_ids":["https://openalex.org/I87182695"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100651384","display_name":"Yanzhi Wang","orcid":"https://orcid.org/0000-0002-3024-7990"},"institutions":[{"id":"https://openalex.org/I87182695","display_name":"Universidad del Noreste","ror":"https://ror.org/02ahky613","country_code":"MX","type":"education","lineage":["https://openalex.org/I87182695"]}],"countries":["MX"],"is_corresponding":false,"raw_author_name":"Yanzhi Wang","raw_affiliation_strings":["Northeastern University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Northeastern University","institution_ids":["https://openalex.org/I87182695"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5073885088","display_name":"Pu Zhao","orcid":"https://orcid.org/0000-0001-5018-2859"},"institutions":[{"id":"https://openalex.org/I87182695","display_name":"Universidad del Noreste","ror":"https://ror.org/02ahky613","country_code":"MX","type":"education","lineage":["https://openalex.org/I87182695"]}],"countries":["MX"],"is_corresponding":false,"raw_author_name":"Pu Zhao","raw_affiliation_strings":["Northeastern University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Northeastern University","institution_ids":["https://openalex.org/I87182695"]}]}],"institutions":[],"countries_distinct_count":3,"institutions_distinct_count":12,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.9349,"has_fulltext":false,"cited_by_count":1,"citation_normalized_percentile":{"value":0.8122501,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":{"min":95,"max":98},"biblio":{"volume":null,"issue":null,"first_page":"1","last_page":"9"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.28060001134872437,"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"}},"topics":[{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.28060001134872437,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.1234000027179718,"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/T14347","display_name":"Big Data and Digital Economy","score":0.10170000046491623,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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":0.6460999846458435},{"id":"https://openalex.org/keywords/security-token","display_name":"Security token","score":0.5758000016212463},{"id":"https://openalex.org/keywords/mobile-device","display_name":"Mobile device","score":0.574400007724762},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.539900004863739},{"id":"https://openalex.org/keywords/edge-device","display_name":"Edge device","score":0.421999990940094},{"id":"https://openalex.org/keywords/cloud-computing","display_name":"Cloud computing","score":0.42149999737739563},{"id":"https://openalex.org/keywords/enhanced-data-rates-for-gsm-evolution","display_name":"Enhanced Data Rates for GSM Evolution","score":0.420199990272522},{"id":"https://openalex.org/keywords/speedup","display_name":"Speedup","score":0.4153999984264374}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.769599974155426},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.6460999846458435},{"id":"https://openalex.org/C48145219","wikidata":"https://www.wikidata.org/wiki/Q1335365","display_name":"Security token","level":2,"score":0.5758000016212463},{"id":"https://openalex.org/C186967261","wikidata":"https://www.wikidata.org/wiki/Q5082128","display_name":"Mobile device","level":2,"score":0.574400007724762},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.539900004863739},{"id":"https://openalex.org/C113775141","wikidata":"https://www.wikidata.org/wiki/Q428691","display_name":"Computer engineering","level":1,"score":0.4253000020980835},{"id":"https://openalex.org/C138236772","wikidata":"https://www.wikidata.org/wiki/Q25098575","display_name":"Edge device","level":3,"score":0.421999990940094},{"id":"https://openalex.org/C79974875","wikidata":"https://www.wikidata.org/wiki/Q483639","display_name":"Cloud computing","level":2,"score":0.42149999737739563},{"id":"https://openalex.org/C162307627","wikidata":"https://www.wikidata.org/wiki/Q204833","display_name":"Enhanced Data Rates for GSM Evolution","level":2,"score":0.420199990272522},{"id":"https://openalex.org/C68339613","wikidata":"https://www.wikidata.org/wiki/Q1549489","display_name":"Speedup","level":2,"score":0.4153999984264374},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.3695000112056732},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.3659000098705292},{"id":"https://openalex.org/C2779960059","wikidata":"https://www.wikidata.org/wiki/Q7113681","display_name":"Overhead (engineering)","level":2,"score":0.336899995803833},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3246999979019165},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.31949999928474426},{"id":"https://openalex.org/C124584101","wikidata":"https://www.wikidata.org/wiki/Q1053266","display_name":"Multiplier (economics)","level":2,"score":0.3156999945640564},{"id":"https://openalex.org/C120314980","wikidata":"https://www.wikidata.org/wiki/Q180634","display_name":"Distributed computing","level":1,"score":0.3091999888420105},{"id":"https://openalex.org/C2779191767","wikidata":"https://www.wikidata.org/wiki/Q6887075","display_name":"Mobile cloud computing","level":3,"score":0.2985000014305115},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.29030001163482666},{"id":"https://openalex.org/C144543869","wikidata":"https://www.wikidata.org/wiki/Q2738570","display_name":"Mobile computing","level":2,"score":0.28940001130104065},{"id":"https://openalex.org/C79403827","wikidata":"https://www.wikidata.org/wiki/Q3988","display_name":"Real-time computing","level":1,"score":0.28290000557899475},{"id":"https://openalex.org/C95491727","wikidata":"https://www.wikidata.org/wiki/Q992968","display_name":"Mobile telephony","level":3,"score":0.27810001373291016},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.26019999384880066},{"id":"https://openalex.org/C127162648","wikidata":"https://www.wikidata.org/wiki/Q16858953","display_name":"Channel (broadcasting)","level":2,"score":0.2583000063896179},{"id":"https://openalex.org/C71901391","wikidata":"https://www.wikidata.org/wiki/Q7126699","display_name":"Upload","level":2,"score":0.25780001282691956},{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.25690001249313354},{"id":"https://openalex.org/C108583219","wikidata":"https://www.wikidata.org/wiki/Q197536","display_name":"Deep learning","level":2,"score":0.2549999952316284}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/iccad66269.2025.11240685","is_oa":false,"landing_page_url":"https://doi.org/10.1109/iccad66269.2025.11240685","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE/ACM International Conference On Computer Aided Design (ICCAD)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":22,"referenced_works":["https://openalex.org/W2045064825","https://openalex.org/W2996728628","https://openalex.org/W3159727696","https://openalex.org/W3202459663","https://openalex.org/W4200630931","https://openalex.org/W4226087293","https://openalex.org/W4312446478","https://openalex.org/W4312910119","https://openalex.org/W4360831786","https://openalex.org/W4386075865","https://openalex.org/W4389519278","https://openalex.org/W4389519600","https://openalex.org/W4393160423","https://openalex.org/W4394593136","https://openalex.org/W4402670692","https://openalex.org/W4403864152","https://openalex.org/W4404781677","https://openalex.org/W4406650295","https://openalex.org/W4409347803","https://openalex.org/W4409363545","https://openalex.org/W4412945475","https://openalex.org/W4413147647"],"related_works":[],"abstract_inverted_index":{"A":[0],"growing":[1],"trend":[2,18],"has":[3],"emerged":[4],"in":[5],"designing":[6],"high-quality":[7],"Small":[8],"Language":[9],"Models":[10],"(SLMs)":[11],"with":[12,69,76,119,207],"a":[13,164,212],"few":[14],"million":[15],"parameters.":[16],"This":[17],"is":[19,36,46,106],"driven":[20],"by":[21,51],"the":[22,96,136,182],"increasing":[23],"concerns":[24],"over":[25],"cloud":[26],"costs,":[27],"privacy,":[28],"and":[29,55,83,131],"latency.":[30],"Considering":[31],"that":[32],"full":[33],"parameter":[34],"training":[35],"feasible":[37],"for":[38,122],"SLMs":[39,102,123],"on":[40,73,88,103,124,158,175],"mobile":[41,82,104,125,176],"devices,":[42,85],"Quantization-Aware":[43],"Training":[44],"(QAT)":[45],"employed":[47],"to":[48,66,101,134,154,170,189,204],"improve":[49],"efficiency":[50],"reducing":[52],"computational":[53],"overhead":[54],"memory":[56],"footprint.":[57],"However,":[58],"previous":[59],"QAT":[60,117,191],"works":[61],"adopt":[62],"fine-grained":[63],"quantization":[64,121],"methods":[65,100,192],"compress":[67],"models":[68],"billions":[70],"of":[71,98,138,185,202],"parameters":[72],"GPUs,":[74],"incompatible":[75],"current":[77],"commodity":[78],"hardware,":[79],"such":[80],"as":[81],"edge":[84],"which":[86],"relies":[87],"Single":[89],"Instruction":[90],"Multiple":[91],"Data":[92],"(SIMD)":[93],"instructions.":[94],"Thus,":[95],"generalization":[97],"these":[99],"devices":[105],"limited.":[107],"In":[108],"this":[109],"paper,":[110],"we":[111,128,144,162,197],"propose":[112,129],"Squat":[113],"method,":[114],"an":[115,199],"effective":[116],"framework":[118],"deployable":[120],"devices.":[126,177],"Specifically,":[127],"entropy-guided":[130],"distribution-aligned":[132],"distillation":[133],"mitigate":[135],"distortion":[137],"attention":[139],"information":[140],"from":[141],"quantization.":[142],"Besides,":[143],"employ":[145],"sub-8-bit":[146,172],"token":[147],"adaptive":[148],"quantization,":[149],"assigning":[150],"varying":[151],"bit":[152],"widths":[153],"different":[155],"tokens":[156],"based":[157],"their":[159],"importance.":[160],"Furthermore,":[161,196],"develop":[163],"SIMD-based":[165],"Multi-Kernel":[166],"Mixed-Precision":[167],"(MKMP)":[168],"multiplier":[169],"support":[171],"mixed-precision":[173],"MAC":[174],"Our":[178],"extensive":[179],"experiments":[180],"verify":[181],"substantial":[183],"improvements":[184],"our":[186],"method":[187],"compared":[188,206],"other":[190],"across":[193],"various":[194],"datasets.":[195],"achieve":[198],"on-device":[200],"speedup":[201],"up":[203],"2.37\u00d7":[205],"its":[208],"FP16":[209],"counterparts,":[210],"signaling":[211],"great":[213],"advancement.":[214],"Code:":[215],"https://github.com/shawnricecake/squant":[216]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-06-11T09:08:48.828518","created_date":"2025-11-20T00:00:00"}