{"id":"https://openalex.org/W4417339299","doi":"https://doi.org/10.1109/ictai66417.2025.00112","title":"Training Generative Judge with Hard Negative Mining: A Metric Learning Perspective","display_name":"Training Generative Judge with Hard Negative Mining: A Metric Learning Perspective","publication_year":2025,"publication_date":"2025-11-03","ids":{"openalex":"https://openalex.org/W4417339299","doi":"https://doi.org/10.1109/ictai66417.2025.00112"},"language":null,"primary_location":{"id":"doi:10.1109/ictai66417.2025.00112","is_oa":false,"landing_page_url":"https://doi.org/10.1109/ictai66417.2025.00112","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE 37th International Conference on Tools with Artificial Intelligence (ICTAI)","raw_type":"proceedings-article"},"type":"article","indexed_in":["crossref"],"open_access":{"is_oa":false,"oa_status":"closed","oa_url":null,"any_repository_has_fulltext":false},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5101917696","display_name":"Changxin Chen","orcid":"https://orcid.org/0000-0003-1468-5920"},"institutions":[{"id":"https://openalex.org/I181679659","display_name":"Shanghai University of Finance and Economics","ror":"https://ror.org/00wtvfq62","country_code":"CN","type":"education","lineage":["https://openalex.org/I181679659"]}],"countries":["CN"],"is_corresponding":true,"raw_author_name":"Changxin Chen","raw_affiliation_strings":["Shanghai University of Finance and Economics,Shanghai,China"],"affiliations":[{"raw_affiliation_string":"Shanghai University of Finance and Economics,Shanghai,China","institution_ids":["https://openalex.org/I181679659"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":["https://openalex.org/A5101917696"],"corresponding_institution_ids":["https://openalex.org/I181679659"],"apc_list":null,"apc_paid":null,"fwci":0.0,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":{"value":0.21035729,"is_in_top_1_percent":false,"is_in_top_10_percent":false},"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":"778","last_page":"785"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.19509999454021454,"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/T10028","display_name":"Topic Modeling","score":0.19509999454021454,"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":0.15139999985694885,"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/T11636","display_name":"Artificial Intelligence in Healthcare and Education","score":0.0754999965429306,"subfield":{"id":"https://openalex.org/subfields/2718","display_name":"Health Informatics"},"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/metric","display_name":"Metric (unit)","score":0.7272999882698059},{"id":"https://openalex.org/keywords/margin","display_name":"Margin (machine learning)","score":0.7171000242233276},{"id":"https://openalex.org/keywords/perspective","display_name":"Perspective (graphical)","score":0.6516000032424927},{"id":"https://openalex.org/keywords/generative-grammar","display_name":"Generative grammar","score":0.6140000224113464},{"id":"https://openalex.org/keywords/context","display_name":"Context (archaeology)","score":0.6098999977111816},{"id":"https://openalex.org/keywords/ranking","display_name":"Ranking (information retrieval)","score":0.5353000164031982},{"id":"https://openalex.org/keywords/training","display_name":"Training (meteorology)","score":0.5108000040054321},{"id":"https://openalex.org/keywords/limit","display_name":"Limit (mathematics)","score":0.44699999690055847},{"id":"https://openalex.org/keywords/generative-model","display_name":"Generative model","score":0.4221000075340271}],"concepts":[{"id":"https://openalex.org/C176217482","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Metric (unit)","level":2,"score":0.7272999882698059},{"id":"https://openalex.org/C774472","wikidata":"https://www.wikidata.org/wiki/Q6760393","display_name":"Margin (machine learning)","level":2,"score":0.7171000242233276},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6664999723434448},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.6516000032424927},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.6442000269889832},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.6140000224113464},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.6098999977111816},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5784000158309937},{"id":"https://openalex.org/C189430467","wikidata":"https://www.wikidata.org/wiki/Q7293293","display_name":"Ranking (information retrieval)","level":2,"score":0.5353000164031982},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.5108000040054321},{"id":"https://openalex.org/C151201525","wikidata":"https://www.wikidata.org/wiki/Q177239","display_name":"Limit (mathematics)","level":2,"score":0.44699999690055847},{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.4221000075340271},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.3652999997138977},{"id":"https://openalex.org/C99844830","wikidata":"https://www.wikidata.org/wiki/Q102441924","display_name":"Scaling","level":2,"score":0.32989999651908875},{"id":"https://openalex.org/C2780586882","wikidata":"https://www.wikidata.org/wiki/Q7520643","display_name":"Simple (philosophy)","level":2,"score":0.31619998812675476},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.3133000135421753},{"id":"https://openalex.org/C112972136","wikidata":"https://www.wikidata.org/wiki/Q7595718","display_name":"Stability (learning theory)","level":2,"score":0.3116999864578247},{"id":"https://openalex.org/C49937458","wikidata":"https://www.wikidata.org/wiki/Q2599292","display_name":"Probabilistic logic","level":2,"score":0.3043000102043152},{"id":"https://openalex.org/C2780898871","wikidata":"https://www.wikidata.org/wiki/Q860554","display_name":"Performance metric","level":2,"score":0.30329999327659607},{"id":"https://openalex.org/C86037889","wikidata":"https://www.wikidata.org/wiki/Q4330127","display_name":"Learning to rank","level":3,"score":0.29649999737739563},{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.2865000069141388},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.2849000096321106},{"id":"https://openalex.org/C204323151","wikidata":"https://www.wikidata.org/wiki/Q905424","display_name":"Range (aeronautics)","level":2,"score":0.28049999475479126},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.2800999879837036},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.26510000228881836},{"id":"https://openalex.org/C13355873","wikidata":"https://www.wikidata.org/wiki/Q2920850","display_name":"Connection (principal bundle)","level":2,"score":0.2646999955177307},{"id":"https://openalex.org/C2776502983","wikidata":"https://www.wikidata.org/wiki/Q690182","display_name":"Contrast (vision)","level":2,"score":0.2639000117778778},{"id":"https://openalex.org/C97931131","wikidata":"https://www.wikidata.org/wiki/Q5282087","display_name":"Discriminative model","level":2,"score":0.25929999351501465}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.1109/ictai66417.2025.00112","is_oa":false,"landing_page_url":"https://doi.org/10.1109/ictai66417.2025.00112","pdf_url":null,"source":null,"license":null,"license_id":null,"version":"publishedVersion","is_accepted":true,"is_published":true,"raw_source_name":"2025 IEEE 37th International Conference on Tools with Artificial Intelligence (ICTAI)","raw_type":"proceedings-article"}],"best_oa_location":null,"sustainable_development_goals":[],"awards":[{"id":"https://openalex.org/G4072165450","display_name":null,"funder_award_id":"25ZR1402137","funder_id":"https://openalex.org/F4320309612","funder_display_name":"Natural Science Foundation of Shanghai"}],"funders":[{"id":"https://openalex.org/F4320309612","display_name":"Natural Science Foundation of Shanghai","ror":null}],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":14,"referenced_works":["https://openalex.org/W1975517671","https://openalex.org/W2096733369","https://openalex.org/W2948638722","https://openalex.org/W2963988212","https://openalex.org/W3034303554","https://openalex.org/W3129831491","https://openalex.org/W3166500718","https://openalex.org/W4385570961","https://openalex.org/W4387321091","https://openalex.org/W4393160410","https://openalex.org/W4401042981","https://openalex.org/W4402671800","https://openalex.org/W4402683797","https://openalex.org/W4409357579"],"related_works":[],"abstract_inverted_index":{"Prior":[0],"SFT-based":[1,148,168],"LLM-as-a-Judge":[2,123],"suffers":[3],"from":[4,176],"a":[5,75,109,126],"distributional":[6],"mismatch":[7],"between":[8],"GPT-4":[9],"and":[10,64,124,142],"the":[11,85,99,120,183],"smaller":[12],"opensource":[13],"language":[14],"models":[15,37],"(SLMs),":[16],"often":[17,51],"leading":[18],"to":[19,31,84,102,147,171,179],"overfitting.":[20],"On-policy":[21],"methods,":[22,149],"such":[23],"as":[24],"Group":[25],"Relative":[26],"Policy":[27],"Optimization":[28],"(GRPO),":[29],"attempt":[30],"address":[32],"this":[33,192],"limitation":[34],"by":[35,78],"training":[36,161,184,205],"using":[38],"their":[39],"own":[40],"outputs.":[41],"However,":[42],"conventional":[43],"GRPO":[44,81,104],"typically":[45],"rely":[46],"on":[47,95,138,162],"randomly":[48],"sampled":[49],"data,":[50,185],"selecting":[52],"negative":[53,107],"samples":[54],"with":[55],"low":[56],"model-assigned":[57],"probabilities-samples":[58],"that":[59,80,196,212],"provide":[60],"weak":[61],"gradient":[62],"signals":[63],"thus":[65],"limit":[66],"learning":[67],"efficacy.":[68],"To":[69],"overcome":[70],"this,":[71,96],"we":[72,97],"first":[73,100],"establish":[74],"theoretical":[76],"connection":[77],"proving":[79],"is":[82,156],"equivalent":[83],"margin":[86],"ranking":[87],"loss":[88],"commonly":[89],"used":[90],"in":[91,112],"metric":[92,113],"learning.":[93,114],"Building":[94],"introduce":[98],"approach":[101,151,214],"optimize":[103],"through":[105],"hard":[106,117],"mining,":[108],"well-established":[110],"technique":[111],"We":[115,134],"define":[116],"negatives":[118],"within":[119],"context":[121],"of":[122,182],"propose":[125],"simple":[127],"yet":[128],"effective":[129],"strategy":[130],"for":[131,160,204],"identifying":[132],"them.":[133],"evaluate":[135],"our":[136,150,186,213],"method":[137,187],"three":[139],"out-of-domain":[140],"datasets":[141],"one":[143],"in-domain":[144],"dataset.":[145],"Compared":[146],"demonstrates":[152],"superior":[153],"scalability,":[154],"which":[155],"becoming":[157],"increasingly":[158],"important":[159],"modern,":[163],"largescale":[164],"datasets.":[165],"Specifically,":[166],"while":[167],"methods":[169,199],"struggle":[170],"improve":[172],"performance":[173],"when":[174],"scaling":[175],"10":[177],"%":[178,181],"100":[180],"exhibits":[188],"consistent":[189],"improvements":[190],"across":[191],"range.":[193],"This":[194],"suggests":[195],"current":[197],"SFT":[198],"may":[200],"not":[201],"be":[202],"well-suited":[203],"generative":[206],"judges.":[207],"Furthermore,":[208],"empirical":[209],"results":[210],"show":[211],"consistently":[215],"outperforms":[216],"larger":[217],"open-source":[218],"baselines.":[219]},"counts_by_year":[],"updated_date":"2026-04-09T08:11:56.329763","created_date":"2025-12-15T00:00:00"}