{"id":"https://openalex.org/W4416037226","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.37","title":"PAFT: Prompt-Agnostic Fine-Tuning","display_name":"PAFT: Prompt-Agnostic Fine-Tuning","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4416037226","doi":"https://doi.org/10.18653/v1/2025.emnlp-main.37"},"language":null,"primary_location":{"id":"doi:10.18653/v1/2025.emnlp-main.37","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.37","pdf_url":"https://aclanthology.org/2025.emnlp-main.37.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":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.emnlp-main.37.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5120309192","display_name":"Chenxing Wei","orcid":null},"institutions":[],"countries":[],"is_corresponding":true,"raw_author_name":"Chenxing Wei","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120309193","display_name":"Yao Shu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yao Shu","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120309194","display_name":"Mingwen Ou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mingwen Ou","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120309195","display_name":"Ying He","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ying He","raw_affiliation_strings":[],"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5120309196","display_name":"Fei Yu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Fei Yu","raw_affiliation_strings":[],"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":5,"corresponding_author_ids":["https://openalex.org/A5120309192"],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":true,"cited_by_count":0,"citation_normalized_percentile":{"value":0.49701735,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"694","last_page":"717"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11930","display_name":"Cardiac, Anesthesia and Surgical Outcomes","score":0.0203000009059906,"subfield":{"id":"https://openalex.org/subfields/2705","display_name":"Cardiology and Cardiovascular Medicine"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}},"topics":[{"id":"https://openalex.org/T11930","display_name":"Cardiac, Anesthesia and Surgical Outcomes","score":0.0203000009059906,"subfield":{"id":"https://openalex.org/subfields/2705","display_name":"Cardiology and Cardiovascular Medicine"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}},{"id":"https://openalex.org/T10372","display_name":"Cardiac Imaging and Diagnostics","score":0.01510000042617321,"subfield":{"id":"https://openalex.org/subfields/2741","display_name":"Radiology, Nuclear Medicine and Imaging"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}},{"id":"https://openalex.org/T11700","display_name":"Hemodynamic Monitoring and Therapy","score":0.010400000028312206,"subfield":{"id":"https://openalex.org/subfields/2746","display_name":"Surgery"},"field":{"id":"https://openalex.org/fields/27","display_name":"Medicine"},"domain":{"id":"https://openalex.org/domains/4","display_name":"Health Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/overfitting","display_name":"Overfitting","score":0.781499981880188},{"id":"https://openalex.org/keywords/generalization","display_name":"Generalization","score":0.6552000045776367},{"id":"https://openalex.org/keywords/robustness","display_name":"Robustness (evolution)","score":0.5200999975204468},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.5034999847412109},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.47429999709129333},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.4643999934196472},{"id":"https://openalex.org/keywords/construct","display_name":"Construct (python library)","score":0.4544999897480011},{"id":"https://openalex.org/keywords/training-set","display_name":"Training set","score":0.43299999833106995}],"concepts":[{"id":"https://openalex.org/C22019652","wikidata":"https://www.wikidata.org/wiki/Q331309","display_name":"Overfitting","level":3,"score":0.781499981880188},{"id":"https://openalex.org/C177148314","wikidata":"https://www.wikidata.org/wiki/Q170084","display_name":"Generalization","level":2,"score":0.6552000045776367},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6517000198364258},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6482999920845032},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.6065000295639038},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.5200999975204468},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.5034999847412109},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.47429999709129333},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.4643999934196472},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.4544999897480011},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.43299999833106995},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.42590001225471497},{"id":"https://openalex.org/C2778334786","wikidata":"https://www.wikidata.org/wiki/Q1586270","display_name":"Variation (astronomy)","level":2,"score":0.34940001368522644},{"id":"https://openalex.org/C117765406","wikidata":"https://www.wikidata.org/wiki/Q5362437","display_name":"Generalization error","level":3,"score":0.3073999881744385},{"id":"https://openalex.org/C197115733","wikidata":"https://www.wikidata.org/wiki/Q1003136","display_name":"Forcing (mathematics)","level":2,"score":0.29649999737739563},{"id":"https://openalex.org/C136389625","wikidata":"https://www.wikidata.org/wiki/Q334384","display_name":"Supervised learning","level":3,"score":0.2913999855518341},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.29120001196861267},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.28769999742507935},{"id":"https://openalex.org/C76969082","wikidata":"https://www.wikidata.org/wiki/Q486902","display_name":"Mathematical model","level":2,"score":0.26809999346733093}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.emnlp-main.37","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.37","pdf_url":"https://aclanthology.org/2025.emnlp-main.37.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":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.emnlp-main.37","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.emnlp-main.37","pdf_url":"https://aclanthology.org/2025.emnlp-main.37.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":"Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G1121271761","display_name":null,"funder_award_id":"Program","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G1231421488","display_name":null,"funder_award_id":"under","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G2030486226","display_name":null,"funder_award_id":"62371309","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G2087396116","display_name":null,"funder_award_id":"China","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G2981938667","display_name":null,"funder_award_id":"Shenzhen","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G3155578932","display_name":null,"funder_award_id":"62271324","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G3317480652","display_name":null,"funder_award_id":"Science","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G3472539505","display_name":null,"funder_award_id":"202205","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G5321231671","display_name":null,"funder_award_id":"6227132","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G5485689011","display_name":null,"funder_award_id":"62231020","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G5939423041","display_name":null,"funder_award_id":"Technology","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G5994120800","display_name":null,"funder_award_id":"Natural","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G7033253288","display_name":null,"funder_award_id":"Grants","funder_id":"https://openalex.org/F4320321001","funder_display_name":"National Natural Science Foundation of China"},{"id":"https://openalex.org/G7473718869","display_name":null,"funder_award_id":"22310","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":{"grobid_xml":true,"pdf":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4416037226.pdf","grobid_xml":"https://content.openalex.org/works/W4416037226.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Fine-tuning":[0],"large":[1],"language":[2],"models":[3,53,113],"(LLMs)":[4],"often":[5],"causes":[6],"overfitting":[7],"to":[8,48,54,93,123],"specific":[9],"prompt":[10,33,79,125],"wording,":[11],"where":[12],"minor":[13],"phrasing":[14],"variations":[15],"drastically":[16],"reduce":[17],"performance.To":[18],"address":[19],"this,":[20],"we":[21],"propose":[22],"Prompt-Agnostic":[23],"Fine-Tuning":[24],"(PAFT),":[25],"a":[26],"method":[27],"that":[28,137],"enhances":[29],"robustness":[30],"through":[31],"dynamic":[32],"variation":[34],"during":[35],"training.PAFT":[36],"first":[37],"generates":[38],"diverse":[39],"synthetic":[40],"prompts,":[41],"then":[42],"continuously":[43],"samples":[44],"from":[45],"this":[46],"set":[47],"construct":[49],"training":[50],"instances,":[51],"forcing":[52],"learn":[55],"fundamental":[56],"task":[57],"principles":[58],"rather":[59],"than":[60,89],"surface-level":[61],"patterns.Across":[62],"systematic":[63],"evaluations":[64],"using":[65],"both":[66],"supervised":[67],"fine-tuning":[68,73],"(SFT)":[69],"and":[70,110],"reinforcement":[71],"learning":[72],"(RLFT),":[74],"PAFT":[75,96,116,138],"demonstrates":[76],"substantially":[77],"improved":[78],"robustness,":[80,95],"achieving":[81],"7%":[82],"higher":[83],"generalization":[84,144],"accuracy":[85],"on":[86,102],"unseen":[87],"prompts":[88],"standard":[90],"methods.In":[91],"addition":[92],"enhanced":[94],"consistently":[97],"yields":[98],"superior":[99],"overall":[100],"performance":[101],"established":[103],"benchmarks":[104],"for":[105],"question":[106],"answering,":[107],"mathematical":[108],"reasoning,":[109],"tool":[111],"use.Notably,":[112],"trained":[114],"with":[115],"attain":[117],"3.2":[118],"faster":[119],"inference":[120],"speeds":[121],"due":[122],"reduced":[124],"sensitivity.Ablation":[126],"studies":[127],"further":[128],"validate":[129],"effectiveness":[130],"of":[131,146],"PAFT,":[132],"while":[133],"theoretical":[134],"analysis":[135],"reveals":[136],"can":[139],"effectively":[140],"enhance":[141],"the":[142],"cross-domain":[143],"ability":[145],"LLM.0":[147],"20":[148],"40":[149],"60":[150],"80":[151],"Accuracy(%)":[152,165],"0.00":[153,166],"0.02":[154],"0.04":[155],"0.06":[156],"0.08":[157],"0.10":[158],"Density":[159],"Winogrande":[160],"25":[161],"50":[162],"75":[163],"100":[164],"0.05":[167],"0.":[168]},"counts_by_year":[],"updated_date":"2026-04-13T07:58:08.660418","created_date":"2025-11-08T00:00:00"}