{"id":"https://openalex.org/W7140133908","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.320","title":"Task-aware Block Pruning with Output Distribution Signals for Large Language Models","display_name":"Task-aware Block Pruning with Output Distribution Signals for Large Language Models","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W7140133908","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.320"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2026.findings-eacl.320","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.320","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.320","pdf_url":"https://aclanthology.org/2026.findings-eacl.320.pdf","source":{"id":"https://openalex.org/S7407087051","display_name":"Findings of the Association for Computational Linguistics: EACL","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"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: EACL 2026","raw_type":"proceedings-article"},"volume":null,"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"hybrid","oa_url":"https://aclanthology.org/2026.findings-eacl.320.pdf","any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5130407203","display_name":"Song-ha Jo","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I139264467","display_name":"Seoul National University","ror":"https://ror.org/04h9pn542","country_code":"KR","type":"education","lineage":["https://openalex.org/I139264467"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Song-ha Jo","raw_affiliation_strings":["Seoul National University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Seoul National University","institution_ids":["https://openalex.org/I139264467"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5110850968","display_name":"Youngrok Ko","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I139264467","display_name":"Seoul National University","ror":"https://ror.org/04h9pn542","country_code":"KR","type":"education","lineage":["https://openalex.org/I139264467"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Youngrok Ko","raw_affiliation_strings":["Seoul National University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Seoul National University","institution_ids":["https://openalex.org/I139264467"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5105936935","display_name":"Sang-goo Lee","orcid":"https://orcid.org/0000-0002-0063-0083","observed_orcids":["https://orcid.org/0000-0002-0063-0083"]},"institutions":[{"id":"https://openalex.org/I139264467","display_name":"Seoul National University","ror":"https://ror.org/04h9pn542","country_code":"KR","type":"education","lineage":["https://openalex.org/I139264467"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Sang-goo Lee","raw_affiliation_strings":["IntelliSys Co. , Ltd","Seoul National University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IntelliSys Co. , Ltd","institution_ids":[]},{"raw_affiliation_string":"Seoul National University","institution_ids":["https://openalex.org/I139264467"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5072440057","display_name":"Jinseok Seol","orcid":"https://orcid.org/0000-0002-7394-6871","observed_orcids":["https://orcid.org/0000-0002-7394-6871"]},"institutions":[{"id":"https://openalex.org/I89015989","display_name":"Dankook University","ror":"https://ror.org/058pdbn81","country_code":"KR","type":"education","lineage":["https://openalex.org/I89015989"]}],"countries":["KR"],"is_corresponding":false,"raw_author_name":"Jinseok Seol","raw_affiliation_strings":["Dankook University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Dankook University","institution_ids":["https://openalex.org/I89015989"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":2,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.14765954,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"6089","last_page":"6107"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10036","display_name":"Advanced Neural Network Applications","score":0.9282000064849854,"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":0.9282000064849854,"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/T10028","display_name":"Topic Modeling","score":0.05169999971985817,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.008999999612569809,"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/large-language-models","display_name":"large language models","score":0.9879999756813049},{"id":"https://openalex.org/keywords/output-distribution","display_name":"output distribution","score":0.6520000100135803},{"id":"https://openalex.org/keywords/inference-latency-reduction","display_name":"inference latency reduction","score":0.7229999899864197},{"id":"https://openalex.org/keywords/model-compression","display_name":"model compression","score":0.628000020980835},{"id":"https://openalex.org/keywords/uncertainty-estimation","display_name":"uncertainty estimation","score":0.6019999980926514},{"id":"https://openalex.org/keywords/calibration-dataset","display_name":"calibration dataset","score":0.6589999794960022},{"id":"https://openalex.org/keywords/transformer-blocks","display_name":"transformer blocks","score":0.3790000081062317}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5478000044822693},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.508400022983551},{"id":"https://openalex.org/C2777210771","wikidata":"https://www.wikidata.org/wiki/Q4927124","display_name":"Block (permutation group theory)","level":2,"score":0.5008000135421753},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.45980000495910645},{"id":"https://openalex.org/C108010975","wikidata":"https://www.wikidata.org/wiki/Q500094","display_name":"Pruning","level":2,"score":0.4287000000476837},{"id":"https://openalex.org/C110121322","wikidata":"https://www.wikidata.org/wiki/Q865811","display_name":"Distribution (mathematics)","level":2,"score":0.4066999852657318},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.383899986743927},{"id":"https://openalex.org/C153180895","wikidata":"https://www.wikidata.org/wiki/Q7148389","display_name":"Pattern recognition (psychology)","level":2,"score":0.36399999260902405},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.30640000104904175},{"id":"https://openalex.org/C149717495","wikidata":"https://www.wikidata.org/wiki/Q117806","display_name":"Gamma distribution","level":2,"score":0.30309998989105225},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.28619998693466187},{"id":"https://openalex.org/C149441793","wikidata":"https://www.wikidata.org/wiki/Q200726","display_name":"Probability distribution","level":2,"score":0.28130000829696655},{"id":"https://openalex.org/C2778112365","wikidata":"https://www.wikidata.org/wiki/Q3511065","display_name":"Sequence (biology)","level":2,"score":0.26660001277923584},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.26649999618530273},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.2547999918460846}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2026.findings-eacl.320","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.320","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.320","pdf_url":"https://aclanthology.org/2026.findings-eacl.320.pdf","source":{"id":"https://openalex.org/S7407087051","display_name":"Findings of the Association for Computational Linguistics: EACL","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"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: EACL 2026","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-eacl.320","doi":"https://doi.org/10.18653/v1/2026.findings-eacl.320","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-eacl.320","pdf_url":"https://aclanthology.org/2026.findings-eacl.320.pdf","source":{"id":"https://openalex.org/S7407087051","display_name":"Findings of the Association for Computational Linguistics: EACL","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"conference"},"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: EACL 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320320671","display_name":"National Research Foundation","ror":"https://ror.org/05s0g1g46"},{"id":"https://openalex.org/F4320321292","display_name":"Seoul National University","ror":"https://ror.org/04h9pn542"},{"id":"https://openalex.org/F4320322120","display_name":"National Research Foundation of Korea","ror":"https://ror.org/013aysd81"},{"id":"https://openalex.org/F4320328359","display_name":"Ministry of Science and ICT, South Korea","ror":"https://ror.org/01wpjm123"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7140133908.pdf","grobid_xml":"https://content.openalex.org/works/W7140133908.grobid-xml"},"referenced_works_count":12,"referenced_works":["https://openalex.org/W2525332836","https://openalex.org/W2896739098","https://openalex.org/W3129585169","https://openalex.org/W4387561528","https://openalex.org/W4400022866","https://openalex.org/W4400761265","https://openalex.org/W4401306886","https://openalex.org/W4414633696","https://openalex.org/W4414883370","https://openalex.org/W4414939355","https://openalex.org/W4416123783","https://openalex.org/W4417089552"],"related_works":["https://openalex.org/W4412888794","https://openalex.org/W7140138985","https://openalex.org/W4416036840","https://openalex.org/W7123403367","https://openalex.org/W6891841569","https://openalex.org/W4416257299","https://openalex.org/W4392270529","https://openalex.org/W6922677330","https://openalex.org/W7138391501","https://openalex.org/W7102329059"],"abstract_inverted_index":{"Large":[0],"language":[1],"models":[2,22],"(LLMs)":[3],"provide":[4,119],"excellent":[5],"performance,":[6],"but":[7],"their":[8],"practical":[9],"deployment":[10],"is":[11],"limited":[12],"by":[13,87],"the":[14,24,35,92,107],"substantial":[15],"compute":[16],"and":[17,23,110,146,157],"memory":[18],"demands":[19],"of":[20,26,37,94,164],"large":[21],"latency":[25,43,145],"auto-regressive":[27],"decoding.To":[28],"mitigate":[29],"these":[30,117],"inefficiencies,":[31],"block":[32,62],"pruning":[33],"reduces":[34],"number":[36],"executed":[38],"transformer":[39],"blocks,":[40],"effectively":[41],"lowering":[42],"while":[44,141],"preserving":[45],"architectural":[46],"coherence.However,":[47],"existing":[48],"methods":[49],"typically":[50],"rely":[51],"on":[52,100,114,151],"representation":[53],"similarity":[54],"or":[55],"computationally":[56],"expensive":[57],"sensitivity":[58,153],"analyses":[59],"to":[60],"estimate":[61],"importance,":[63],"thereby":[64],"neglecting":[65],"task-aware":[66],"model":[67],"behavior.To":[68],"address":[69],"this":[70],"limitation,":[71],"we":[72,160],"introduce":[73],"Task-aware":[74],"Block":[75],"Pruning":[76],"(TaBP),":[77],"a":[78,101,120],"novel":[79],"approach":[80],"that":[81,126,135],"directly":[82],"captures":[83],"task-specific":[84],"inference":[85,144],"dynamics":[86],"quantifying":[88],"block-level":[89],"uncertainty":[90,112],"from":[91],"statistics":[93,118],"each":[95],"block's":[96],"early-exited":[97],"output":[98,104],"distribution":[99],"calibration":[102],"dataset.Since":[103],"distributions":[105],"reflect":[106],"model's":[108],"confidence":[109],"decision":[111],"conditioned":[113],"downstream":[115,138],"tasks,":[116],"principled":[121],"signal":[122],"for":[123,130],"identifying":[124],"blocks":[125],"are":[127],"less":[128],"critical":[129],"task":[131,139],"performance.Extensive":[132],"experiments":[133],"demonstrate":[134],"TaBP":[136],"preserves":[137],"performance":[140],"substantially":[142],"reducing":[143],"computational":[147],"cost,":[148],"without":[149],"relying":[150],"cost-heavy":[152],"analyses.To":[154],"facilitate":[155],"reproducibility":[156],"further":[158],"research,":[159],"release":[161],"our":[162],"implementation":[163],"TaBP.":[165]},"counts_by_year":[],"updated_date":"2026-11-05T12:25:35.769452","created_date":"2026-03-24T00:00:00"}