{"id":"https://openalex.org/W4416034562","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.645","title":"Revisiting Pruning vs Quantization for Small Language Models","display_name":"Revisiting Pruning vs Quantization for Small Language Models","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416034562","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.645"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.645","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.645","pdf_url":"https://aclanthology.org/2025.findings-emnlp.645.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.findings-emnlp.645.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101957983","display_name":"Zihan Zhou","orcid":"https://orcid.org/0009-0002-9375-1905"},"institutions":[{"id":"https://openalex.org/I200332995","display_name":"TU Dortmund University","ror":"https://ror.org/01k97gp34","country_code":"DE","type":"education","lineage":["https://openalex.org/I200332995"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Zihan Zhou","raw_affiliation_strings":["Tsinghua University TU Dortmund University , Lamarr Institute University of Sheffield"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University TU Dortmund University , Lamarr Institute University of Sheffield","institution_ids":["https://openalex.org/I200332995"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5108632299","display_name":"S. Kurz","orcid":null},"institutions":[{"id":"https://openalex.org/I200332995","display_name":"TU Dortmund University","ror":"https://ror.org/01k97gp34","country_code":"DE","type":"education","lineage":["https://openalex.org/I200332995"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Simon Kurz","raw_affiliation_strings":["Tsinghua University TU Dortmund University , Lamarr Institute University of Sheffield"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University TU Dortmund University , Lamarr Institute University of Sheffield","institution_ids":["https://openalex.org/I200332995"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5101991159","display_name":"Zhixue Zhao","orcid":"https://orcid.org/0000-0002-3060-269X"},"institutions":[{"id":"https://openalex.org/I200332995","display_name":"TU Dortmund University","ror":"https://ror.org/01k97gp34","country_code":"DE","type":"education","lineage":["https://openalex.org/I200332995"]}],"countries":["DE"],"is_corresponding":false,"raw_author_name":"Zhixue Zhao","raw_affiliation_strings":["Tsinghua University TU Dortmund University , Lamarr Institute University of Sheffield"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tsinghua University TU Dortmund University , Lamarr Institute University of Sheffield","institution_ids":["https://openalex.org/I200332995"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I200332995"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"12055","last_page":"12070"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.42660000920295715,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.42660000920295715,"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/T10201","display_name":"Speech Recognition and Synthesis","score":0.07020000368356705,"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/T10028","display_name":"Topic Modeling","score":0.04839999973773956,"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 (signal processing)","score":0.5055000185966492},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.39719998836517334},{"id":"https://openalex.org/keywords/pruning","display_name":"Pruning","score":0.3359000086784363},{"id":"https://openalex.org/keywords/natural-language","display_name":"Natural language","score":0.3116999864578247},{"id":"https://openalex.org/keywords/feature","display_name":"Feature (linguistics)","score":0.2904999852180481}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5403000116348267},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5160999894142151},{"id":"https://openalex.org/C28855332","wikidata":"https://www.wikidata.org/wiki/Q198099","display_name":"Quantization (signal processing)","level":2,"score":0.5055000185966492},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.39719998836517334},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.39480000734329224},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.3781999945640564},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.3668000102043152},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.3359000086784363},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.3116999864578247},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.2904999852180481},{"id":"https://openalex.org/C199833920","wikidata":"https://www.wikidata.org/wiki/Q612536","display_name":"Vector quantization","level":2,"score":0.2863999903202057},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.27869999408721924},{"id":"https://openalex.org/C28490314","wikidata":"https://www.wikidata.org/wiki/Q189436","display_name":"Speech recognition","level":1,"score":0.2529999911785126}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-emnlp.645","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.645","pdf_url":"https://aclanthology.org/2025.findings-emnlp.645.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.645","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.645","pdf_url":"https://aclanthology.org/2025.findings-emnlp.645.pdf","source":null,"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"Findings of the Association for Computational Linguistics: EMNLP 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416034562.pdf","grobid_xml":"https://content.openalex.org/works/W4416034562.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Deploying":[0],"language":[1],"models":[2,18],"on":[3,88,135],"resourceconstrained":[4],"devices,":[5],"such":[6],"as":[7],"mobile":[8],"phones,":[9],"wearables,":[10],"and":[11,51,65,82,91,102,131],"on-device":[12],"AI":[13],"assistants,":[14],"demands":[15],"compact,":[16],"efficient":[17],"without":[19],"sacrificing":[20],"performance.Compressing":[21],"Small":[22],"Language":[23,41],"Models":[24,42],"(SLMs)":[25],"is":[26],"particularly":[27],"suited":[28],"for":[29,128],"these":[30],"scenarios,":[31],"yet":[32],"their":[33],"compression":[34,100,130],"dynamics":[35],"remain":[36],"underexplored":[37],"compared":[38],"to":[39,61,114,119],"Large":[40],"(LLMs).We":[43],"systematically":[44],"evaluate":[45],"leading":[46],"post-training":[47],"pruning":[48,75],"(SparseGPT,":[49],"Wanda)":[50],"quantization":[52,72,125],"(GPTQ,":[53],"AWQ)":[54,127],"methods":[55],"across":[56],"six":[57],"SLMs":[58],"from":[59],"0.5":[60],"3.8B,":[62],"seven":[63,66],"languages,":[64],"downstream":[67,103],"tasks.Our":[68],"results":[69],"show":[70],"that":[71],"consistently":[73],"outperforms":[74],"in":[76,108],"preserving":[77],"model":[78],"fidelity,":[79],"multilingual":[80],"perplexity,":[81],"reasoning":[83,92],"accuracy.However,":[84],"quantization's":[85],"advantages":[86],"diminish":[87],"complex":[89],"knowledge":[90],"tasks":[93],"like":[94],"OpenBookQA,":[95],"highlighting":[96],"a":[97,136],"disconnect":[98],"between":[99],"fidelity":[101],"task":[104],"performance.Notably,":[105],"trends":[106],"observed":[107],"LLMs":[109],"(e.g.,":[110],"Wanda's":[111],"competitive":[112],"performance":[113],"SparseGPT)":[115],"do":[116],"not":[117],"generalize":[118],"SLMs.For":[120],"practitioners,":[121],"we":[122],"recommend":[123],"prioritizing":[124],"(particularly":[126],"SLM":[129],"caution":[132],"against":[133],"relying":[134],"single":[137],"metric.":[138]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-11-08T00:00:00"}