{"id":"https://openalex.org/W4417258219","doi":"https://doi.org/10.18653/v1/2026.findings-acl.713","title":"Towards Bridging the Reward-Generation Gap in Direct Alignment Algorithms","display_name":"Towards Bridging the Reward-Generation Gap in Direct Alignment Algorithms","publication_year":2026,"publication_date":"2026-01-01","ids":{"openalex":"https://openalex.org/W4417258219","doi":"https://doi.org/10.18653/v1/2026.findings-acl.713"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2026.findings-acl.713","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.713","pdf_url":"https://aclanthology.org/2026.findings-acl.713.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 2026","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["arxiv","crossref","datacite"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2026.findings-acl.713.pdf","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":null,"display_name":"Zeguan Xiao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zeguan Xiao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100416547","display_name":"Yun Chen","orcid":"https://orcid.org/0000-0002-4988-8894"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yun Chen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Jian Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jian Yang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100665989","display_name":"Guanhua Chen","orcid":"https://orcid.org/0000-0003-2917-7566"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guanhua Chen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5021254405","display_name":"Ke Tang","orcid":"https://orcid.org/0000-0002-6236-2002"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ke Tang","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":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.03385999,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"14534","last_page":"14548"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T12535","display_name":"Machine Learning and Data Classification","score":0.2304999977350235,"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/T12535","display_name":"Machine Learning and Data Classification","score":0.2304999977350235,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.12319999933242798,"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/T10203","display_name":"Recommender Systems and Techniques","score":0.08550000190734863,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/bridging","display_name":"Bridging (networking)","score":0.7164999842643738},{"id":"https://openalex.org/keywords/perspective","display_name":"Perspective (graphical)","score":0.5192000269889832},{"id":"https://openalex.org/keywords/process","display_name":"Process (computing)","score":0.5054000020027161},{"id":"https://openalex.org/keywords/optimization-algorithm","display_name":"Optimization algorithm","score":0.46540001034736633},{"id":"https://openalex.org/keywords/global-optimization","display_name":"Global optimization","score":0.44429999589920044},{"id":"https://openalex.org/keywords/simple","display_name":"Simple (philosophy)","score":0.43779999017715454},{"id":"https://openalex.org/keywords/preference","display_name":"Preference","score":0.41940000653266907},{"id":"https://openalex.org/keywords/optimization-problem","display_name":"Optimization problem","score":0.4171000123023987}],"concepts":[{"id":"https://openalex.org/C174348530","wikidata":"https://www.wikidata.org/wiki/Q188635","display_name":"Bridging (networking)","level":2,"score":0.7164999842643738},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6916999816894531},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.5192000269889832},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.5054000020027161},{"id":"https://openalex.org/C2987595161","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Optimization algorithm","level":2,"score":0.46540001034736633},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.4645000100135803},{"id":"https://openalex.org/C164752517","wikidata":"https://www.wikidata.org/wiki/Q5570875","display_name":"Global optimization","level":2,"score":0.44429999589920044},{"id":"https://openalex.org/C2780586882","wikidata":"https://www.wikidata.org/wiki/Q7520643","display_name":"Simple (philosophy)","level":2,"score":0.43779999017715454},{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.41940000653266907},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.4171000123023987},{"id":"https://openalex.org/C141603448","wikidata":"https://www.wikidata.org/wiki/Q134830","display_name":"Prefix","level":2,"score":0.38429999351501465},{"id":"https://openalex.org/C68781425","wikidata":"https://www.wikidata.org/wiki/Q2052203","display_name":"Multi-objective optimization","level":2,"score":0.3749000132083893},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.3659999966621399},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.3508000075817108},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3443000018596649},{"id":"https://openalex.org/C100776233","wikidata":"https://www.wikidata.org/wiki/Q2532492","display_name":"Bridge (graph theory)","level":2,"score":0.33959999680519104},{"id":"https://openalex.org/C126255220","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Mathematical optimization","level":1,"score":0.3246999979019165},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3156000077724457},{"id":"https://openalex.org/C55660270","wikidata":"https://www.wikidata.org/wiki/Q5164377","display_name":"Constrained optimization","level":2,"score":0.3068999946117401},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.2953999936580658},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.2653000056743622}],"mesh":[],"locations_count":3,"locations":[{"id":"doi:10.18653/v1/2026.findings-acl.713","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.713","pdf_url":"https://aclanthology.org/2026.findings-acl.713.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 2026","raw_type":"proceedings-article"},{"id":"pmh:oai:arXiv.org:2506.09457","is_oa":true,"landing_page_url":"https://arxiv.org/abs/2506.09457","pdf_url":"https://arxiv.org/pdf/2506.09457","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,"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2506.09457","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2506.09457","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,"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.18653/v1/2026.findings-acl.713","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2026.findings-acl.713","pdf_url":"https://aclanthology.org/2026.findings-acl.713.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 2026","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[{"id":"https://openalex.org/F4320309612","display_name":"Natural Science Foundation of Shanghai","ror":null},{"id":"https://openalex.org/F4320321001","display_name":"National Natural Science Foundation of China","ror":"https://ror.org/01h0zpd94"},{"id":"https://openalex.org/F4320327605","display_name":"Southern University of Science and Technology","ror":null},{"id":"https://openalex.org/F4320337111","display_name":"Basic and Applied Basic Research Foundation of Guangdong Province","ror":null}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4417258219.pdf","grobid_xml":"https://content.openalex.org/works/W4417258219.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Direct":[0,6],"Alignment":[1],"Algorithms":[2],"(DAAs),":[3],"such":[4],"as":[5,17,45],"Preference":[7,12],"Optimization":[8,13],"(DPO)":[9],"and":[10,53,83,111,127,140,161],"Simple":[11],"(SimPO),":[14],"have":[15],"emerged":[16],"efficient":[18],"alternatives":[19],"to":[20,64,107,130,155,171],"Reinforcement":[21],"Learning":[22],"from":[23,39],"Human":[24],"Feedback":[25],"(RLHF)":[26],"algorithms":[27],"for":[28],"aligning":[29,180],"large":[30],"language":[31],"models":[32],"(LLMs)":[33],"with":[34,138,183],"human":[35],"preferences.However,":[36],"DAAs":[37,106,177],"suffer":[38],"a":[40,101,113],"fundamental":[41],"limitation":[42],"we":[43,59,99],"identify":[44],"the":[46,65,69,72,79,90,97,132,169,173],"\"reward-generation":[47],"gap\"-a":[48],"discrepancy":[49],"between":[50,71],"training":[51,181],"objectives":[52,182],"autoregressive":[54,184],"decoding":[55,185],"dynamics.In":[56],"this":[57,85],"paper,":[58],"consider":[60],"that":[61,146],"one":[62],"contributor":[63],"reward-generation":[66,174],"gap":[67,175],"is":[68,87],"mismatch":[70],"inherent":[73],"importance":[74,86],"of":[75,94,105],"prefix":[76],"tokens":[77],"during":[78],"LLM":[80],"generation":[81],"process":[82],"how":[84],"reflected":[88],"in":[89,158,176],"implicit":[91],"reward":[92],"functions":[93],"DAAs.To":[95],"bridge":[96],"gap,":[98],"adopt":[100],"token-level":[102],"MDP":[103],"perspective":[104],"analyze":[108],"its":[109],"limitations":[110],"introduce":[112],"simple":[114],"yet":[115],"effective":[116],"approach":[117],"called":[118],"Prefix-Oriented":[119],"Equal-length":[120],"Training":[121],"(POET),":[122],"which":[123],"truncates":[124],"both":[125],"preferred":[126],"dispreferred":[128],"responses":[129],"match":[131],"shorter":[133],"one's":[134],"length.We":[135],"conduct":[136],"experiments":[137],"DPO":[139],"SimPO,":[141],"two":[142],"representative":[143],"DAAs,":[144],"demonstrating":[145],"POET":[147],"improves":[148],"over":[149],"their":[150],"standard":[151],"implementations,":[152],"achieving":[153],"up":[154],"11.8":[156],"points":[157],"AlpacaEval":[159],"2":[160],"overall":[162],"improvements":[163],"across":[164],"downstream":[165],"tasks.These":[166],"results":[167],"underscore":[168],"need":[170],"mitigate":[172],"by":[178],"better":[179],"dynamics.":[186]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2025-10-10T00:00:00"}