{"id":"https://openalex.org/W4416034230","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.951","title":"Beyond Binary Preferences: Semi-Online Label-Free GRACE-KTO with Group-Wise Adaptive Calibration for High-Quality Long-Text Generation","display_name":"Beyond Binary Preferences: Semi-Online Label-Free GRACE-KTO with Group-Wise Adaptive Calibration for High-Quality Long-Text Generation","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416034230","doi":"https://doi.org/10.18653/v1/2025.findings-emnlp.951"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.951","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.951","pdf_url":"https://aclanthology.org/2025.findings-emnlp.951.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: EMNLP 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-emnlp.951.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5102565860","display_name":"Jingyang Deng","orcid":"https://orcid.org/0009-0008-3487-225X"},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jingyang Deng","raw_affiliation_strings":["School of Mathematical Sciences and LMAM , Peking University , Beijing 100871 , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Mathematical Sciences and LMAM , Peking University , Beijing 100871 , China","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100400794","display_name":"Ran Chen","orcid":"https://orcid.org/0000-0001-7876-5546"},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Ran Chen","raw_affiliation_strings":["School of Mathematical Sciences and LMAM , Peking University , Beijing 100871 , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Mathematical Sciences and LMAM , Peking University , Beijing 100871 , China","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5013467571","display_name":"Jo-Ku Cheng","orcid":null},"institutions":[{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jo-Ku Cheng","raw_affiliation_strings":["School of Mathematical Sciences and LMAM , Peking University , Beijing 100871 , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"School of Mathematical Sciences and LMAM , Peking University , Beijing 100871 , China","institution_ids":["https://openalex.org/I20231570"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5108638166","display_name":"Jinwen Ma","orcid":"https://orcid.org/0009-0006-5861-1957"},"institutions":[{"id":"https://openalex.org/I176432857","display_name":"Beijing Wuzi University","ror":"https://ror.org/00bd1d647","country_code":"CN","type":"education","lineage":["https://openalex.org/I176432857"]},{"id":"https://openalex.org/I20231570","display_name":"Peking University","ror":"https://ror.org/02v51f717","country_code":"CN","type":"education","lineage":["https://openalex.org/I20231570"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Jinwen Ma","raw_affiliation_strings":["Institute of Systems Science , Beijing Wuzi University , Beijing 101149 , China","School of Mathematical Sciences and LMAM , Peking University , Beijing 100871 , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Institute of Systems Science , Beijing Wuzi University , Beijing 101149 , China","institution_ids":["https://openalex.org/I176432857"]},{"raw_affiliation_string":"School of Mathematical Sciences and LMAM , Peking University , Beijing 100871 , China","institution_ids":["https://openalex.org/I20231570"]}]}],"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":0.28917088,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"17550","last_page":"17562"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11550","display_name":"Text and Document Classification Technologies","score":0.14509999752044678,"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/T11550","display_name":"Text and Document Classification Technologies","score":0.14509999752044678,"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":0.1307000070810318,"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":0.0869000032544136,"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/calibration","display_name":"Calibration","score":0.6036999821662903},{"id":"https://openalex.org/keywords/binary-number","display_name":"Binary number","score":0.47870001196861267},{"id":"https://openalex.org/keywords/binary-data","display_name":"Binary data","score":0.3779999911785126},{"id":"https://openalex.org/keywords/noise","display_name":"Noise (video)","score":0.30730000138282776}],"concepts":[{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.6036999821662903},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5008000135421753},{"id":"https://openalex.org/C48372109","wikidata":"https://www.wikidata.org/wiki/Q3913","display_name":"Binary number","level":2,"score":0.47870001196861267},{"id":"https://openalex.org/C11413529","wikidata":"https://www.wikidata.org/wiki/Q8366","display_name":"Algorithm","level":1,"score":0.42579999566078186},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.41600000858306885},{"id":"https://openalex.org/C2779190172","wikidata":"https://www.wikidata.org/wiki/Q4913888","display_name":"Binary data","level":3,"score":0.3779999911785126},{"id":"https://openalex.org/C33923547","wikidata":"https://www.wikidata.org/wiki/Q395","display_name":"Mathematics","level":0,"score":0.3499999940395355},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.30730000138282776},{"id":"https://openalex.org/C2780009758","wikidata":"https://www.wikidata.org/wiki/Q6804172","display_name":"Measure (data warehouse)","level":2,"score":0.27149999141693115},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.27079999446868896}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-emnlp.951","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.951","pdf_url":"https://aclanthology.org/2025.findings-emnlp.951.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: EMNLP 2025","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.findings-emnlp.951","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-emnlp.951","pdf_url":"https://aclanthology.org/2025.findings-emnlp.951.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: EMNLP 2025","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G2870131413","display_name":null,"funder_award_id":"62471175","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"},{"id":"https://openalex.org/F4320324787","display_name":"Peking University","ror":"https://ror.org/02v51f717"}],"has_content":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416034230.pdf","grobid_xml":"https://content.openalex.org/works/W4416034230.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Generating":[0],"high-quality":[1],"long-text":[2,177],"remains":[3],"challenging":[4],"for":[5,40],"Large":[6],"Language":[7],"Models":[8],"(LLMs),":[9],"as":[10,26],"conventional":[11],"supervised":[12],"fine-tuning":[13],"fails":[14],"to":[15,20,81,100,110,135,146,156],"ensure":[16],"overall":[17],"quality":[18,58,92],"due":[19],"its":[21,137],"teacherforcing":[22],"nature.Kahneman-Tversky":[23],"Optimization":[24],"(KTO),":[25],"a":[27,65,106],"model":[28,155],"alignment":[29],"method":[30],"that":[31,54,68],"can":[32],"holistically":[33],"optimize":[34],"generation":[35,121,139,178],"quality,":[36],"overcomes":[37],"the":[38,131,143,154],"need":[39],"paired":[41],"preference":[42],"data":[43,128],"required":[44],"by":[45],"previous":[46],"methods.However,":[47],"it":[48],"still":[49],"suffers":[50],"from":[51],"binary":[52,71],"supervision":[53],"inadequately":[55],"reflects":[56],"varying":[57],"degrees.To":[59],"address":[60],"this,":[61],"we":[62,125,141],"propose":[63],"GRACE-KTO,":[64],"semionline":[66,107],"framework":[67],"transforms":[69],"KTO's":[70],"signals":[72],"into":[73,84],"dynamically":[74],"calibrated":[75],"intra-group":[76],"rewards.Specifically,":[77],"GRACE-KTO":[78],"aggregates":[79],"responses":[80],"identical":[82],"queries":[83],"groups,":[85],"computes":[86],"rank-sum":[87],"scores":[88],"across":[89],"multiple":[90],"linguistic":[91],"dimensions,":[93],"and":[94,97,173],"applies":[95],"groupwise":[96],"global":[98],"normalization":[99],"adaptively":[101],"redistribute":[102],"sample":[103],"importance.We":[104],"adopt":[105],"training":[108],"strategy":[109],"reduce":[111],"costly":[112],"online":[113],"sampling":[114],"while":[115,160],"outperforming":[116],"offline":[117],"variants.By":[118],"leveraging":[119],"query":[120],"with":[122],"seed":[123],"data,":[124],"minimize":[126],"labeled":[127],"dependency,":[129],"using":[130,149],"model's":[132],"own":[133],"knowledge":[134],"enhance":[136],"longtext":[138],"capabilities.Additionally,":[140],"extend":[142],"context":[144],"window":[145],"32k":[147],"tokens":[148],"YaRN":[150],"during":[151],"inference,":[152],"enabling":[153],"generate":[157],"longer":[158],"texts":[159],"maintaining":[161],"perplexities.Experiments":[162],"demonstrate":[163],"GRACE-KTO's":[164],"superiority":[165],"over":[166],"vanilla":[167],"KTO":[168],"on":[169],"both":[170],"automatic":[171],"metrics":[172],"LLM-as-a-Judge":[174],"evaluations,":[175],"advancing":[176],"through":[179],"group-wise":[180],"adaptive":[181],"calibration.":[182]},"counts_by_year":[],"updated_date":"2026-08-22T07:34:49.880490","created_date":"2025-11-08T00:00:00"}