{"id":"https://openalex.org/W7125939770","doi":"https://doi.org/10.1109/smc58881.2025.11342731","title":"SmoothRot: Combining Channel-Wise Scaling and Rotation for Quantization-Friendly LLMs","display_name":"SmoothRot: Combining Channel-Wise Scaling and Rotation for Quantization-Friendly LLMs","publication_year":2025,"publication_date":"2025-10-05","ids":{"openalex":"https://openalex.org/W7125939770","doi":"https://doi.org/10.1109/smc58881.2025.11342731"},"language":null,"primary_location":{"id":"doi:10.1109/smc58881.2025.11342731","is_oa":false,"landing_page_url":"https://doi.org/10.1109/smc58881.2025.11342731","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE International Conference on Systems, Man, and Cybernetics (SMC)","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/A5117490015","display_name":"Patrik Czak\u00f3","orcid":null},"institutions":[{"id":"https://openalex.org/I103356709","display_name":"Obuda University","ror":"https://ror.org/00ax71d21","country_code":"HU","type":"education","lineage":["https://openalex.org/I103356709"]}],"countries":["HU"],"is_corresponding":false,"raw_author_name":"Patrik Czak\u00f3","raw_affiliation_strings":["Obuda University,Doctoral School of Applied Informatics and Applied Mathematics,Budapest,Hungary"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Obuda University,Doctoral School of Applied Informatics and Applied Mathematics,Budapest,Hungary","institution_ids":["https://openalex.org/I103356709"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5084426852","display_name":"G\u00e1bor Kert\u00e9sz","orcid":"https://orcid.org/0000-0002-8845-8301"},"institutions":[{"id":"https://openalex.org/I103356709","display_name":"Obuda University","ror":"https://ror.org/00ax71d21","country_code":"HU","type":"education","lineage":["https://openalex.org/I103356709"]}],"countries":["HU"],"is_corresponding":false,"raw_author_name":"G\u00e1bor Kert\u00e9sz","raw_affiliation_strings":["Obuda University,John von Neumann Faculty of Informatics,Budapest,Hungary"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Obuda University,John von Neumann Faculty of Informatics,Budapest,Hungary","institution_ids":["https://openalex.org/I103356709"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5024487424","display_name":"S\u00e1ndor Sz\u00e9n\u00e1si","orcid":"https://orcid.org/0000-0002-7292-0717"},"institutions":[{"id":"https://openalex.org/I103356709","display_name":"Obuda University","ror":"https://ror.org/00ax71d21","country_code":"HU","type":"education","lineage":["https://openalex.org/I103356709"]}],"countries":["HU"],"is_corresponding":false,"raw_author_name":"S\u00e1ndor Sz\u00e9n\u00e1si","raw_affiliation_strings":["Obuda University,John von Neumann Faculty of Informatics,Budapest,Hungary"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Obuda University,John von Neumann Faculty of Informatics,Budapest,Hungary","institution_ids":["https://openalex.org/I103356709"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I103356709"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.65613843,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"6461","last_page":"6466"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.2378000020980835,"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/T10028","display_name":"Topic Modeling","score":0.2378000020980835,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.12960000336170197,"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.11219999939203262,"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":0.8425999879837036},{"id":"https://openalex.org/keywords/scaling","display_name":"Scaling","score":0.763700008392334},{"id":"https://openalex.org/keywords/hadamard-transform","display_name":"Hadamard transform","score":0.6589000225067139},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5320000052452087},{"id":"https://openalex.org/keywords/outlier","display_name":"Outlier","score":0.5099999904632568},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.37540000677108765}],"concepts":[{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.8425999879837036},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.763700008392334},{"id":"https://openalex.org/C60292330","wikidata":"https://www.wikidata.org/wiki/Q1014065","display_name":"Hadamard transform","level":2,"score":0.6589000225067139},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5320000052452087},{"id":"https://openalex.org/C79337645","wikidata":"https://www.wikidata.org/wiki/Q779824","display_name":"Outlier","level":2,"score":0.5099999904632568},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.46860000491142273},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.4341999888420105},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.37540000677108765},{"id":"https://openalex.org/C74050887","wikidata":"https://www.wikidata.org/wiki/Q848368","display_name":"Rotation (mathematics)","level":2,"score":0.3626999855041504},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3305000066757202},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.30889999866485596},{"id":"https://openalex.org/C2776760102","wikidata":"https://www.wikidata.org/wiki/Q5139990","display_name":"Code (set theory)","level":3,"score":0.29280000925064087},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.28060001134872437},{"id":"https://openalex.org/C149782125","wikidata":"https://www.wikidata.org/wiki/Q160039","display_name":"Econometrics","level":1,"score":0.262800008058548},{"id":"https://openalex.org/C97256817","wikidata":"https://www.wikidata.org/wiki/Q1462316","display_name":"Spurious relationship","level":2,"score":0.2565999925136566},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.2558000087738037}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/smc58881.2025.11342731","is_oa":false,"landing_page_url":"https://doi.org/10.1109/smc58881.2025.11342731","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE International Conference on Systems, Man, and Cybernetics (SMC)","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":12,"referenced_works":["https://openalex.org/W1632114991","https://openalex.org/W2946609015","https://openalex.org/W2963015836","https://openalex.org/W2963122961","https://openalex.org/W2998617917","https://openalex.org/W3194676777","https://openalex.org/W4394831376","https://openalex.org/W4404782932","https://openalex.org/W4410226559","https://openalex.org/W4411552739","https://openalex.org/W4415796210","https://openalex.org/W4415796640"],"related_works":[],"abstract_inverted_index":{"We":[0],"present":[1],"SmoothRot,":[2],"a":[3],"novel":[4],"post-training":[5],"quantization":[6,14,47],"technique":[7,37],"to":[8],"enhance":[9],"the":[10,22,66],"efficiency":[11],"of":[12,25],"4-bit":[13],"in":[15],"Large":[16],"Language":[17],"Models":[18],"(LLMs).":[19],"SmoothRot":[20,63],"addresses":[21],"critical":[23],"challenge":[24],"massive":[26],"activation":[27],"outliers,":[28],"by":[29,74],"integrating":[30],"channel-wise":[31],"scaling":[32],"with":[33],"Hadamard":[34],"transformations.":[35],"Our":[36],"effectively":[38],"transforms":[39],"extreme":[40],"outliers":[41],"into":[42],"quantization-friendly":[43],"activations,":[44],"significantly":[45],"improving":[46],"accuracy.":[48],"Experiments":[49],"conducted":[50],"on":[51],"popular":[52],"LLMs":[53],"(LLaMA2":[54],"7B,":[55],"LLaMA3.1":[56],"8B,":[57],"and":[58,71,80],"Mistral":[59],"7B)":[60],"demonstrate":[61],"that":[62],"consistently":[64],"reduces":[65],"performance":[67],"gap":[68],"between":[69],"quantized":[70],"FP16":[72],"models":[73],"approximately":[75],"10-30%":[76],"across":[77],"language":[78],"generation":[79],"zero-shot":[81],"reasoning":[82],"tasks,":[83],"without":[84],"introducing":[85],"additional":[86],"inference":[87],"latency.":[88],"Code":[89],"is":[90],"available":[91],"at":[92],"https://github.com/czakop/smoothrot":[93]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2026-01-29T00:00:00"}