{"id":"https://openalex.org/W4412887809","doi":"https://doi.org/10.18653/v1/2025.findings-acl.984","title":"Sens-Merging: Sensitivity-Guided Parameter Balancing for Merging Large Language Models","display_name":"Sens-Merging: Sensitivity-Guided Parameter Balancing for Merging Large Language Models","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412887809","doi":"https://doi.org/10.18653/v1/2025.findings-acl.984"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.984","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.984","pdf_url":"https://aclanthology.org/2025.findings-acl.984.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: ACL 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-acl.984.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101578217","display_name":"Shuqi Liu","orcid":"https://orcid.org/0000-0003-0092-3529"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shuqi Liu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5052351065","display_name":"Han Wu","orcid":"https://orcid.org/0000-0002-2128-2786"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Han Wu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5079956921","display_name":"Bowei He","orcid":"https://orcid.org/0000-0002-0360-2950"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bowei He","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100693994","display_name":"Xiuping Han","orcid":"https://orcid.org/0000-0002-8501-3497"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xiongwei Han","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5109233250","display_name":"Mingxuan Yuan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mingxuan Yuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5101305536","display_name":"Linqi Song","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Linqi Song","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":1,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"19243","last_page":"19255"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.9840999841690063,"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":0.9840999841690063,"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":0.9753999710083008,"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/T10456","display_name":"Multi-Agent Systems and Negotiation","score":0.9660000205039978,"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/computer-science","display_name":"Computer science","score":0.7549885511398315},{"id":"https://openalex.org/keywords/sensitivity","display_name":"Sensitivity (control systems)","score":0.6660376787185669},{"id":"https://openalex.org/keywords/theoretical-computer-science","display_name":"Theoretical computer science","score":0.32317501306533813},{"id":"https://openalex.org/keywords/engineering","display_name":"Engineering","score":0.09036552906036377}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7549885511398315},{"id":"https://openalex.org/C21200559","wikidata":"https://www.wikidata.org/wiki/Q7451068","display_name":"Sensitivity (control systems)","level":2,"score":0.6660376787185669},{"id":"https://openalex.org/C80444323","wikidata":"https://www.wikidata.org/wiki/Q2878974","display_name":"Theoretical computer science","level":1,"score":0.32317501306533813},{"id":"https://openalex.org/C127413603","wikidata":"https://www.wikidata.org/wiki/Q11023","display_name":"Engineering","level":0,"score":0.09036552906036377},{"id":"https://openalex.org/C24326235","wikidata":"https://www.wikidata.org/wiki/Q126095","display_name":"Electronic engineering","level":1,"score":0.0}],"mesh":[],"locations_count":2,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.984","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.984","pdf_url":"https://aclanthology.org/2025.findings-acl.984.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: ACL 2025","raw_type":"proceedings-article"},{"id":"pmh:oai:pure.atira.dk:publications/3d8cdf27-c81d-4aa3-aa8e-3a0121e499d7","is_oa":true,"landing_page_url":"https://hdl.handle.net/2031/3d8cdf27-c81d-4aa3-aa8e-3a0121e499d7","pdf_url":null,"source":{"id":"https://openalex.org/S7407055387","display_name":"CityU Scholars","issn_l":null,"issn":null,"is_oa":false,"is_in_doaj":false,"is_core":false,"host_organization":null,"host_organization_name":null,"host_organization_lineage":[],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"info:eu-repo/semantics/conferenceObject"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-acl.984","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.984","pdf_url":"https://aclanthology.org/2025.findings-acl.984.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: ACL 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/W4412887809.pdf","grobid_xml":"https://content.openalex.org/works/W4412887809.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W4391375266","https://openalex.org/W2899084033","https://openalex.org/W2748952813","https://openalex.org/W2390279801","https://openalex.org/W4391913857","https://openalex.org/W2358668433","https://openalex.org/W4396701345","https://openalex.org/W2376932109","https://openalex.org/W2001405890","https://openalex.org/W4396696052"],"abstract_inverted_index":{"Recent":[0],"advances":[1],"in":[2,129],"large":[3],"language":[4],"models":[5,96,122],"have":[6],"led":[7],"to":[8,85,123],"numerous":[9],"task-specialized":[10],"fine-tuned":[11,126],"variants,":[12],"creating":[13],"a":[14,55],"need":[15],"for":[16,144],"efficient":[17],"model":[18,63,146],"merging":[19,32,64,88,116,147],"techniques":[20,65],"that":[21,60,98],"preserve":[22],"specialized":[23,125],"capabilities":[24],"while":[25],"avoiding":[26],"costly":[27],"retraining.While":[28],"existing":[29,62,115],"task":[30],"vector-based":[31],"methods":[33],"show":[34],"promise,":[35],"they":[36],"typically":[37],"apply":[38],"uniform":[39],"coefficients":[40],"across":[41,51,103],"all":[42],"parameters,":[43],"overlooking":[44],"varying":[45],"parameter":[46,76],"importance":[47],"both":[48,69],"within":[49,78],"and":[50,71,81,94,108,139],"tasks.We":[52],"present":[53],"Sens-Merging,":[54],"sensitivity-guided":[56],"coefficient":[57],"adjustment":[58],"method":[59,74,119],"enhances":[61],"by":[66],"operating":[67],"at":[68],"taskspecific":[70],"cross-task":[72,83,140],"levels.Our":[73],"analyzes":[75],"sensitivity":[77],"individual":[79],"tasks":[80],"evaluates":[82],"transferability":[84],"determine":[86],"optimal":[87],"coefficients.Extensive":[89],"experiments":[90],"on":[91],"Mistral":[92],"7B":[93],"LLaMA2-7B/13B":[95],"demonstrate":[97],"Sens-Merging":[99],"significantly":[100],"improves":[101],"performance":[102],"general":[104],"knowledge,":[105],"mathematical":[106],"reasoning,":[107],"code":[109,130],"generation":[110,131],"tasks.Notably,":[111],"when":[112],"combined":[113],"with":[114],"techniques,":[117],"our":[118],"enables":[120],"merged":[121],"outperform":[124],"models,":[127],"particularly":[128],"tasks.Our":[132],"findings":[133],"reveal":[134],"important":[135],"trade-offs":[136],"between":[137],"task-specific":[138],"scalings,":[141],"providing":[142],"insights":[143],"future":[145],"strategies.":[148]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-07-19T07:52:34.831488","created_date":"2025-10-10T00:00:00"}