{"id":"https://openalex.org/W4412887720","doi":"https://doi.org/10.18653/v1/2025.findings-acl.1078","title":"Disentangling Reasoning Tokens and Boilerplate Tokens For Language Model Fine-tuning","display_name":"Disentangling Reasoning Tokens and Boilerplate Tokens For Language Model Fine-tuning","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412887720","doi":"https://doi.org/10.18653/v1/2025.findings-acl.1078"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.findings-acl.1078","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.1078","pdf_url":"https://aclanthology.org/2025.findings-acl.1078.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.1078.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5020855034","display_name":"Zezhen Ye","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ziang Ye","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5032572801","display_name":"Zhenru Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhenru Zhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100354708","display_name":"Yang Zhang","orcid":"https://orcid.org/0000-0002-6804-1840"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang Zhang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100661619","display_name":"Jianxin Ma","orcid":"https://orcid.org/0009-0000-5214-432X"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Jianxin Ma","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":null,"display_name":"Junyang Lin","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Junyang Lin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5051925942","display_name":"Fuli Feng","orcid":"https://orcid.org/0000-0002-5828-9842"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fuli Feng","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":0.14192427,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"20939","last_page":"20957"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10181","display_name":"Natural Language Processing Techniques","score":0.9794999957084656,"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/T10181","display_name":"Natural Language Processing Techniques","score":0.9794999957084656,"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":0.9070000052452087,"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/boilerplate-text","display_name":"Boilerplate text","score":0.9079738259315491},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.7301514148712158},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.45338523387908936},{"id":"https://openalex.org/keywords/natural-language-processing","display_name":"Natural language processing","score":0.37295451760292053},{"id":"https://openalex.org/keywords/programming-language","display_name":"Programming language","score":0.2560299038887024}],"concepts":[{"id":"https://openalex.org/C75701414","wikidata":"https://www.wikidata.org/wiki/Q1651672","display_name":"Boilerplate text","level":2,"score":0.9079738259315491},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7301514148712158},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.45338523387908936},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.37295451760292053},{"id":"https://openalex.org/C199360897","wikidata":"https://www.wikidata.org/wiki/Q9143","display_name":"Programming language","level":1,"score":0.2560299038887024}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.findings-acl.1078","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.1078","pdf_url":"https://aclanthology.org/2025.findings-acl.1078.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.1078","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.findings-acl.1078","pdf_url":"https://aclanthology.org/2025.findings-acl.1078.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":[{"display_name":"Reduced inequalities","id":"https://metadata.un.org/sdg/10","score":0.6700000166893005}],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412887720.pdf","grobid_xml":"https://content.openalex.org/works/W4412887720.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/W3121258117","https://openalex.org/W4379160565","https://openalex.org/W2146642953","https://openalex.org/W2507103705","https://openalex.org/W4210431233","https://openalex.org/W50124057","https://openalex.org/W2250653840"],"abstract_inverted_index":{"When":[0],"using":[1],"agent-task":[2],"datasets":[3],"to":[4,81],"enhance":[5],"agent":[6],"capabilities":[7],"for":[8,61],"Large":[9],"Language":[10],"Models":[11],"(LLMs),":[12],"current":[13],"methodologies":[14],"often":[15],"treat":[16],"all":[17],"tokens":[18,26,31,34,66,91,106],"within":[19],"a":[20,56],"sample":[21],"equally.However,":[22],"we":[23,54,95],"argue":[24],"that":[25],"serving":[27],"different":[28],"roles-specifically,":[29],"reasoning":[30,90,105],"versus":[32],"boilerplate":[33,78],"(e.g.,":[35],"those":[36],"governing":[37],"output":[38],"format)-differ":[39],"significantly":[40],"in":[41],"importance":[42],"and":[43,49],"learning":[44],"complexity,":[45],"necessitating":[46],"their":[47,82],"disentanglement":[48],"distinct":[50],"treatment.To":[51],"address":[52],"this,":[53],"propose":[55,96],"novel":[57],"Shuffle-Aware":[58],"Discriminator":[59],"(SHAD)":[60],"adaptive":[62],"token":[63],"discrimination.SHAD":[64],"classifies":[65],"by":[67],"exploiting":[68],"predictability":[69],"differences":[70],"observed":[71],"after":[72],"shuffling":[73],"input-output":[74],"combinations":[75],"across":[76],"samples:":[77],"tokens,":[79],"due":[80],"repetitive":[83],"nature":[84],"among":[85],"samples,":[86],"maintain":[87],"predictability,":[88],"whereas":[89],"do":[92],"not.Using":[93],"SHAD,":[94],"the":[97],"Reasoning-highlighted":[98],"Fine-Tuning":[99,116],"(RFT)":[100],"method,":[101],"which":[102],"adaptively":[103],"emphasizes":[104],"during":[107],"fine-tuning,":[108],"yielding":[109],"notable":[110],"performance":[111],"gains":[112],"over":[113],"common":[114],"Supervised":[115],"(SFT).":[117]},"counts_by_year":[],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2025-10-10T00:00:00"}