{"id":"https://openalex.org/W4412887859","doi":"https://doi.org/10.18653/v1/2025.findings-acl.1057","title":"MWPO: Enhancing LLMs Performance through Multi-Weight Preference Strength and Length Optimization","display_name":"MWPO: Enhancing LLMs Performance through Multi-Weight Preference Strength and Length Optimization","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412887859","doi":"https://doi.org/10.18653/v1/2025.findings-acl.1057"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.1057","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.1057","pdf_url":"https://aclanthology.org/2025.findings-acl.1057.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.1057.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5085626484","display_name":"Shutao Xu","orcid":"https://orcid.org/0000-0003-4722-8371"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shiyue Xu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100460011","display_name":"Fu Zhang","orcid":"https://orcid.org/0000-0002-8636-4168"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fu Zhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5052002380","display_name":"Jingwei Cheng","orcid":"https://orcid.org/0000-0003-1054-4952"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jingwei Cheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5085593848","display_name":"Li Zhou","orcid":"https://orcid.org/0000-0003-1945-3491"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Linfeng Zhou","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":"20566","last_page":"20581"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11814","display_name":"Advanced Manufacturing and Logistics Optimization","score":0.9003000259399414,"subfield":{"id":"https://openalex.org/subfields/2209","display_name":"Industrial and Manufacturing Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T11814","display_name":"Advanced Manufacturing and Logistics Optimization","score":0.9003000259399414,"subfield":{"id":"https://openalex.org/subfields/2209","display_name":"Industrial and Manufacturing Engineering"},"field":{"id":"https://openalex.org/fields/22","display_name":"Engineering"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/preference","display_name":"Preference","score":0.5924252867698669},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.3803219199180603},{"id":"https://openalex.org/keywords/business","display_name":"Business","score":0.3324574828147888},{"id":"https://openalex.org/keywords/economics","display_name":"Economics","score":0.2752726078033447},{"id":"https://openalex.org/keywords/microeconomics","display_name":"Microeconomics","score":0.22495123744010925}],"concepts":[{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.5924252867698669},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.3803219199180603},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.3324574828147888},{"id":"https://openalex.org/C162324750","wikidata":"https://www.wikidata.org/wiki/Q8134","display_name":"Economics","level":0,"score":0.2752726078033447},{"id":"https://openalex.org/C175444787","wikidata":"https://www.wikidata.org/wiki/Q39072","display_name":"Microeconomics","level":1,"score":0.22495123744010925}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.1057","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.1057","pdf_url":"https://aclanthology.org/2025.findings-acl.1057.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"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-acl.1057","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.1057","pdf_url":"https://aclanthology.org/2025.findings-acl.1057.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":[{"id":"https://openalex.org/G2799773603","display_name":null,"funder_award_id":"62276057","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"}],"funders":[{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"}],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412887859.pdf","grobid_xml":"https://content.openalex.org/works/W4412887859.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":{"Direct":[0],"Preference":[1],"Optimization":[2],"(DPO)":[3],"have":[4,149],"proposed":[5],"offline":[6],"alternatives":[7],"to":[8,34,42,90,110,129,148,185],"Reinforcement":[9],"Learning":[10],"from":[11,68],"Human":[12],"Feedback":[13],"(RLHF).In":[14],"DPO,":[15],"each":[16],"preference":[17,55,85,101,112,137,141],"pair,":[18],"which":[19,61,87],"serves":[20],"as":[21],"the":[22,35,44,48,53,69,79,197],"foundation":[23],"for":[24,73,133],"learning,":[25],"is":[26,78],"typically":[27],"constructed":[28],"by":[29,183,194],"first":[30],"generating":[31],"multiple":[32,174],"responses":[33,49],"same":[36],"instruction":[37],"and":[38,103,119],"then":[39],"annotating":[40],"them":[41,124],"indicate":[43],"preferred":[45],"choice.However,":[46],"when":[47],"are":[50],"highly":[51],"similar,":[52],"weak":[54],"signal":[56],"can":[57,88],"introduce":[58],"annotation":[59],"noise,":[60],"may":[62],"hinder":[63],"model":[64,155],"optimization.Additionally,":[65],"DPO":[66,182],"suffers":[67],"drawback":[70],"of":[71,81,169],"over-optimizing":[72],"verbose":[74],"generation.A":[75],"potential":[76],"reason":[77],"presence":[80],"length":[82,91,104,121,146,193],"bias":[83],"in":[84,196],"datasets,":[86],"lead":[89],"exploitation.To":[92],"address":[93],"these":[94],"issues,":[95],"we":[96,108],"propose":[97,109],"a":[98,126,150],"DPO-based":[99],"multi-weight":[100],"strength":[102],"optimization":[105],"(MWPO)":[106],"method.Specifically,":[107],"reweight":[111],"pairs":[113,138],"based":[114],"on":[115,154,167,187],"implicit":[116],"reward":[117],"margins":[118],"response":[120],"margins,":[122],"unifying":[123],"through":[125],"geometric":[127],"mixture":[128],"generate":[130],"synthetic":[131],"weights":[132],"optimization.This":[134],"method":[135,158,166,176],"allows":[136],"with":[139],"stronger":[140],"signals":[142],"or":[143],"more":[144,151],"favorable":[145],"feature":[147],"pronounced":[152],"impact":[153],"parameters.Moreover,":[156],"our":[157,165],"does":[159],"not":[160],"require":[161],"additional":[162],"annotators.We":[163],"validate":[164],"models":[168],"four":[170],"different":[171],"scales":[172],"across":[173],"benchmarks.Our":[175],"surpasses":[177],"state-of-the-art":[178],"(SOTA)":[179],"baselines,":[180],"outperforming":[181],"up":[184],"8.7%":[186],"AlpacaEval":[188],"2":[189],"while":[190],"reducing":[191],"generation":[192],"9.4%":[195],"Mistral":[198],"setting":[199],"1":[200],".":[201]},"counts_by_year":[],"updated_date":"2026-07-14T23:27:15.235271","created_date":"2025-10-10T00:00:00"}