{"id":"https://openalex.org/W7214975403","doi":"https://doi.org/10.48550/arxiv.2609.36032","title":"TORQUE: Optimizing What (not) to Quantize Before and After Rotation","display_name":"TORQUE: Optimizing What (not) to Quantize Before and After Rotation","publication_year":2026,"publication_date":"2026-09-28","ids":{"openalex":"https://openalex.org/W7214975403","doi":"https://doi.org/10.48550/arxiv.2609.36032"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2609.36032","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.36032","pdf_url":"https://arxiv.org/pdf/2609.36032","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2609.36032","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5000201729","display_name":"Ran Ben Basat","orcid":"https://orcid.org/0000-0003-0196-9190","observed_orcids":["https://orcid.org/0000-0003-0196-9190"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Basat, Ran Ben","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5045507561","display_name":"Michael Mitzenmacher","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mitzenmacher, Michael","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5078049483","display_name":"Shay Vargaftik","orcid":"https://orcid.org/0000-0002-0982-7894","observed_orcids":["https://orcid.org/0000-0002-0982-7894"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Vargaftik, Shay","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":true,"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/T10901","display_name":"Advanced Data Compression Techniques","score":0.7128999829292297,"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/T10901","display_name":"Advanced Data Compression Techniques","score":0.7128999829292297,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.11990000307559967,"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/T11612","display_name":"Stochastic Gradient Optimization Techniques","score":0.020400000736117363,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/quantization","display_name":"quantization","score":0.5920000076293945},{"id":"https://openalex.org/keywords/random-rotation","display_name":"random rotation","score":0.6129999756813049},{"id":"https://openalex.org/keywords/coordinate-selection","display_name":"coordinate selection","score":0.5360000133514404},{"id":"https://openalex.org/keywords/normal-distribution","display_name":"normal distribution","score":0.5299999713897705},{"id":"https://openalex.org/keywords/bit-budget","display_name":"bit budget","score":0.5640000104904175},{"id":"https://openalex.org/keywords/nearest-neighbor-search","display_name":"nearest neighbor search","score":0.7590000033378601},{"id":"https://openalex.org/keywords/kv-cache-compression","display_name":"KV cache compression","score":0.984000027179718},{"id":"https://openalex.org/keywords/activation-compression","display_name":"activation compression","score":0.9710000157356262}],"concepts":[{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.6700000166893005},{"id":"https://openalex.org/C163716315","wikidata":"https://www.wikidata.org/wiki/Q901177","display_name":"Gaussian","level":2,"score":0.6606000065803528},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.5896000266075134},{"id":"https://openalex.org/C74050887","wikidata":"https://www.wikidata.org/wiki/Q848368","display_name":"Rotation (mathematics)","level":2,"score":0.5769000053405762},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.49380001425743103},{"id":"https://openalex.org/C34736171","wikidata":"https://www.wikidata.org/wiki/Q918333","display_name":"Preprocessor","level":2,"score":0.46970000863075256},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4237000048160553},{"id":"https://openalex.org/C77553402","wikidata":"https://www.wikidata.org/wiki/Q13222579","display_name":"Upper and lower bounds","level":2,"score":0.41499999165534973},{"id":"https://openalex.org/C149629883","wikidata":"https://www.wikidata.org/wiki/Q660926","display_name":"Fraction (chemistry)","level":2,"score":0.4043000042438507},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.3671000003814697},{"id":"https://openalex.org/C61326573","wikidata":"https://www.wikidata.org/wiki/Q1496376","display_name":"Gaussian process","level":3,"score":0.3569999933242798},{"id":"https://openalex.org/C81917197","wikidata":"https://www.wikidata.org/wiki/Q628760","display_name":"Selection (genetic algorithm)","level":2,"score":0.3375999927520752},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.3034000098705292},{"id":"https://openalex.org/C7218915","wikidata":"https://www.wikidata.org/wiki/Q1054475","display_name":"Gaussian function","level":3,"score":0.3012000024318695},{"id":"https://openalex.org/C122383733","wikidata":"https://www.wikidata.org/wiki/Q865920","display_name":"Approximation error","level":2,"score":0.2858000099658966},{"id":"https://openalex.org/C51267290","wikidata":"https://www.wikidata.org/wiki/Q5527848","display_name":"Gaussian random field","level":4,"score":0.2854999899864197},{"id":"https://openalex.org/C80551277","wikidata":"https://www.wikidata.org/wiki/Q11210","display_name":"Coordinate system","level":2,"score":0.2662999927997589},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.25099998712539673},{"id":"https://openalex.org/C126990412","wikidata":"https://www.wikidata.org/wiki/Q203218","display_name":"Spherical coordinate system","level":2,"score":0.2502000033855438}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2609.36032","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.36032","pdf_url":"https://arxiv.org/pdf/2609.36032","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2609.36032","doi":"https://doi.org/10.48550/arxiv.2609.36032","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2609.36032","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2609.36032","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2609.36032","pdf_url":"https://arxiv.org/pdf/2609.36032","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7214975403.pdf","grobid_xml":"https://content.openalex.org/works/W7214975403.grobid-xml"},"referenced_works_count":24,"referenced_works":["https://openalex.org/W2124509324","https://openalex.org/W3092262909","https://openalex.org/W4393147284","https://openalex.org/W4399175194","https://openalex.org/W4404239922","https://openalex.org/W4404240312","https://openalex.org/W4405655184","https://openalex.org/W4411119700","https://openalex.org/W4411403229","https://openalex.org/W4413756934","https://openalex.org/W4415240458","https://openalex.org/W4415445364","https://openalex.org/W4415796210","https://openalex.org/W4415796420","https://openalex.org/W7117651193","https://openalex.org/W7127094184","https://openalex.org/W7133195237","https://openalex.org/W7133198810","https://openalex.org/W7155108335","https://openalex.org/W7157259268","https://openalex.org/W7161144170","https://openalex.org/W7167713981","https://openalex.org/W7201865557","https://openalex.org/W7203861970"],"related_works":["https://openalex.org/W7160596758","https://openalex.org/W7143427932","https://openalex.org/W4399115604","https://openalex.org/W7147406933","https://openalex.org/W4412889461","https://openalex.org/W7161972079","https://openalex.org/W7127120426","https://openalex.org/W4323570450","https://openalex.org/W7161144170","https://openalex.org/W7214734185"],"abstract_inverted_index":{"Uniform":[0],"random":[1,37],"rotations":[2,38],"are":[3],"an":[4,148,169],"effective":[5],"preprocessing":[6],"step":[7],"for":[8,116,137,164],"quantization:":[9],"they":[10],"make":[11],"normalized":[12],"coordinate":[13,145],"distributions":[14],"approximately":[15],"Gaussian,":[16],"enabling":[17,152],"the":[18,80,97,104,117,142,182],"use":[19,36],"of":[20,96],"codebooks":[21,113,159],"optimized":[22,114],"offline.":[23],"We":[24,122,167],"introduce":[25],"TORQUE,":[26],"a":[27,58,93,124,153],"framework":[28],"that":[29,35,131,156],"improves":[30],"on":[31,187],"previous":[32],"quantization":[33,125],"works":[34],"by":[39,78],"jointly":[40],"optimizing":[41],"how":[42],"many":[43,87],"and":[44,54,129,160,175,185,192],"which":[45],"coordinates":[46,70,99],"to":[47,107,147],"preserve":[48],"at":[49,71,100],"high":[50,72,101],"precision":[51,73,102],"both":[52],"before":[53,65],"after":[55,90],"rotation,":[56,66,91],"under":[57,181],"fixed":[59],"overall":[60,76],"expected":[61],"bit":[62],"budget.":[63],"Intuitively,":[64],"preserving":[67,92],"large":[68],"input":[69],"can":[74],"reduce":[75],"error":[77,126],"preventing":[79],"rotation":[81],"from":[82],"spreading":[83],"their":[84],"values":[85,106],"across":[86],"coordinates.":[88],"Likewise,":[89],"small":[94],"fraction":[95],"largest-magnitude":[98],"allows":[103],"remaining":[105],"be":[108],"quantized":[109],"more":[110],"accurately":[111],"using":[112],"offline":[115,158],"resulting":[118],"truncated":[119],"Gaussian":[120,183],"distribution.":[121],"derive":[123],"upper":[127],"bound":[128],"prove":[130],"top-$k$":[132],"pre-rotation":[133],"retention":[134],"minimizes":[135],"it":[136],"each":[138],"$k$.":[139],"This":[140],"reduces":[141],"search":[143],"over":[144,150],"subsets":[146],"optimization":[149],"$k$,":[151],"fast":[154],"optimizer":[155],"uses":[157],"parallel":[161],"parameter":[162],"selection":[163],"practical":[165],"implementation.":[166],"demonstrate":[168],"improved":[170],"tradeoff":[171],"between":[172],"reconstruction":[173],"accuracy":[174],"storage":[176],"cost":[177],"through":[178],"numerical":[179],"evaluation":[180],"model":[184],"experiments":[186],"nearest-neighbor":[188],"retrieval,":[189],"KV-cache":[190],"compression,":[191],"activation":[193],"compression.":[194]},"counts_by_year":[],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2026-10-01T00:00:00"}