{"id":"https://openalex.org/W4403344767","doi":"https://doi.org/10.48550/arxiv.2410.06293","title":"Accelerated Preference Optimization for Large Language Model Alignment","display_name":"Accelerated Preference Optimization for Large Language Model Alignment","publication_year":2024,"publication_date":"2024-10-08","ids":{"openalex":"https://openalex.org/W4403344767","doi":"https://doi.org/10.48550/arxiv.2410.06293"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2410.06293","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2410.06293","pdf_url":"https://arxiv.org/pdf/2410.06293","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2410.06293","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5081773147","display_name":"Jie Ying He","orcid":"https://orcid.org/0000-0002-7136-4181","observed_orcids":["https://orcid.org/0000-0002-7136-4181"]},"institutions":[{"id":"https://openalex.org/I161318765","display_name":"University of California, Los Angeles","ror":"https://ror.org/046rm7j60","country_code":"US","type":"education","lineage":["https://openalex.org/I161318765"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"He, Jiafan","raw_affiliation_strings":["Department of Computer Science , University of California , Los Angeles , CA 90095 , USA;"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science , University of California , Los Angeles , CA 90095 , USA;","institution_ids":["https://openalex.org/I161318765"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5013438281","display_name":"Huizhuo Yuan","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I161318765","display_name":"University of California, Los Angeles","ror":"https://ror.org/046rm7j60","country_code":"US","type":"education","lineage":["https://openalex.org/I161318765"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Yuan, Huizhuo","raw_affiliation_strings":["Department of Computer Science , University of California , Los Angeles , CA 90095 , USA;"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science , University of California , Los Angeles , CA 90095 , USA;","institution_ids":["https://openalex.org/I161318765"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5051448391","display_name":"Quanquan Gu","orcid":"https://orcid.org/0000-0001-9830-793X","observed_orcids":["https://orcid.org/0000-0001-9830-793X"]},"institutions":[{"id":"https://openalex.org/I161318765","display_name":"University of California, Los Angeles","ror":"https://ror.org/046rm7j60","country_code":"US","type":"education","lineage":["https://openalex.org/I161318765"]}],"countries":["US"],"is_corresponding":false,"raw_author_name":"Gu, Quanquan","raw_affiliation_strings":["Department of Computer Science , University of California , Los Angeles , CA 90095 , USA;"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Department of Computer Science , University of California , Los Angeles , CA 90095 , USA;","institution_ids":["https://openalex.org/I161318765"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I161318765"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":5,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10462","display_name":"Reinforcement Learning in Robotics","score":0.5698000192642212,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.5698000192642212,"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":0.11659999936819077,"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":0.09449999779462814,"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-model-alignment","display_name":"large language model alignment","score":0.777999997138977},{"id":"https://openalex.org/keywords/reinforcement-learning-from-human-feedback","display_name":"reinforcement learning from human feedback","score":0.9150000214576721},{"id":"https://openalex.org/keywords/preference-optimization","display_name":"preference optimization","score":0.7879999876022339},{"id":"https://openalex.org/keywords/accelerated-optimization","display_name":"accelerated optimization","score":0.8809999823570251},{"id":"https://openalex.org/keywords/proximal-point-algorithm","display_name":"proximal point algorithm","score":0.949999988079071},{"id":"https://openalex.org/keywords/direct-preference-optimization","display_name":"direct preference optimization","score":0.8690000176429749},{"id":"https://openalex.org/keywords/convergence-rate","display_name":"convergence rate","score":0.7850000262260437}],"concepts":[{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.6761071681976318},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.49953150749206543},{"id":"https://openalex.org/C162324750","wikidata":"https://www.wikidata.org/wiki/Q8134","display_name":"Economics","level":0,"score":0.24274849891662598},{"id":"https://openalex.org/C175444787","wikidata":"https://www.wikidata.org/wiki/Q39072","display_name":"Microeconomics","level":1,"score":0.14570388197898865}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2410.06293","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2410.06293","pdf_url":"https://arxiv.org/pdf/2410.06293","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2410.06293","doi":"https://doi.org/10.48550/arxiv.2410.06293","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2410.06293","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2410.06293","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2410.06293","pdf_url":"https://arxiv.org/pdf/2410.06293","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4403344767.pdf","grobid_xml":"https://content.openalex.org/works/W4403344767.grobid-xml"},"referenced_works_count":31,"referenced_works":["https://openalex.org/W1860742233","https://openalex.org/W2153749284","https://openalex.org/W2626639528","https://openalex.org/W2736601468","https://openalex.org/W2973379954","https://openalex.org/W3043226849","https://openalex.org/W3100569850","https://openalex.org/W3177813494","https://openalex.org/W3192231545","https://openalex.org/W3195140297","https://openalex.org/W3210904070","https://openalex.org/W4223908421","https://openalex.org/W4327810158","https://openalex.org/W4360836968","https://openalex.org/W4377121468","https://openalex.org/W4384918448","https://openalex.org/W4387322976","https://openalex.org/W4387561528","https://openalex.org/W4389470731","https://openalex.org/W4390437503","https://openalex.org/W4390573302","https://openalex.org/W4390723302","https://openalex.org/W4391047180","https://openalex.org/W4391555989","https://openalex.org/W4392616695","https://openalex.org/W4394591101","https://openalex.org/W4396650433","https://openalex.org/W4412886848","https://openalex.org/W6959939111","https://openalex.org/W7019882953","https://openalex.org/W7133208539"],"related_works":["https://openalex.org/W4403576238","https://openalex.org/W7108937016","https://openalex.org/W4401042987","https://openalex.org/W4392121102","https://openalex.org/W4394906585","https://openalex.org/W6910759220","https://openalex.org/W4416043113","https://openalex.org/W6929507215","https://openalex.org/W4409686497","https://openalex.org/W7139079365"],"abstract_inverted_index":{"Reinforcement":[0],"Learning":[1],"from":[2],"Human":[3],"Feedback":[4],"(RLHF)":[5],"has":[6],"emerged":[7],"as":[8,33,125],"a":[9,34,93,126,136,169],"pivotal":[10],"tool":[11],"for":[12,202],"aligning":[13],"large":[14],"language":[15],"models":[16],"(LLMs)":[17],"with":[18],"human":[19],"preferences.":[20],"Direct":[21],"Preference":[22,139,184],"Optimization":[23,140,185],"(DPO),":[24],"one":[25],"of":[26,51,160,192],"the":[27,41,46,59,65,109,117,158,174,190,205],"most":[28],"popular":[29],"approaches,":[30,53],"formulates":[31],"RLHF":[32,73,98,203],"policy":[35,66,69],"optimization":[36,70,77,88,120,148,178],"problem":[37],"without":[38],"explicitly":[39],"estimating":[40,58],"reward":[42,60],"function.":[43],"It":[44],"overcomes":[45],"stability":[47],"and":[48,62,79,91,150,182,198],"efficiency":[49],"issues":[50],"two-step":[52],"which":[54,143],"typically":[55],"involve":[56],"first":[57,114],"function":[61],"then":[63],"optimizing":[64],"via":[67],"proximal":[68,127],"(PPO).":[71],"Since":[72],"is":[74,81],"essentially":[75],"an":[76],"problem,":[78],"it":[80],"well-known":[82],"that":[83,116,165],"momentum":[84,153],"techniques":[85],"can":[86,122,167],"accelerate":[87],"both":[89],"theoretically":[90],"empirically,":[92],"natural":[94],"question":[95,107],"arises:":[96],"Can":[97],"be":[99,123],"accelerated":[100],"by":[101],"momentum?":[102],"This":[103],"paper":[104],"answers":[105],"this":[106,132],"in":[108],"affirmative.":[110],"In":[111],"detail,":[112],"we":[113,134,163,188],"show":[115,189],"iterative":[118,176,196],"preference":[119,147,177],"method":[121],"viewed":[124],"point":[128],"method.":[129],"Based":[130],"on":[131,204],"observation,":[133],"propose":[135],"general":[137],"Accelerated":[138],"(APO)":[141],"framework,":[142],"unifies":[144],"many":[145],"existing":[146],"algorithms":[149],"employs":[151],"Nesterov's":[152],"technique":[154],"to":[155],"speed":[156],"up":[157],"alignment":[159],"LLMs.":[161],"Theoretically,":[162],"demonstrate":[164],"APO":[166,193],"achieve":[168],"faster":[170],"convergence":[171],"rate":[172],"than":[173],"standard":[175],"methods,":[179],"including":[180],"DPO":[181],"Self-Play":[183],"(SPPO).":[186],"Empirically,":[187],"superiority":[191],"over":[194],"DPO,":[195,197],"other":[199],"strong":[200],"baselines":[201],"AlpacaEval":[206],"2.0":[207],"benchmark.":[208]},"counts_by_year":[{"year":2026,"cited_by_count":3},{"year":2025,"cited_by_count":2}],"updated_date":"2026-11-07T11:25:07.637180","created_date":"2024-10-12T00:00:00"}