{"id":"https://openalex.org/W4412886854","doi":"https://doi.org/10.18653/v1/2025.acl-long.34","title":"JuStRank: Benchmarking LLM Judges for System Ranking","display_name":"JuStRank: Benchmarking LLM Judges for System Ranking","publication_year":2025,"publication_date":"2025-01-01","ids":{"openalex":"https://openalex.org/W4412886854","doi":"https://doi.org/10.18653/v1/2025.acl-long.34"},"language":"en","primary_location":{"id":"doi:10.18653/v1/2025.acl-long.34","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.34","pdf_url":"https://aclanthology.org/2025.acl-long.34.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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"type":"conference-paper","indexed_in":["crossref"],"open_access":{"is_oa":true,"oa_status":"gold","oa_url":"https://aclanthology.org/2025.acl-long.34.pdf","any_repository_has_fulltext":null},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5009715645","display_name":"Ariel Gera","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ariel Gera","raw_affiliation_strings":["IBM Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IBM Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5006234414","display_name":"Odellia Boni","orcid":"https://orcid.org/0000-0001-8637-4860"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Odellia Boni","raw_affiliation_strings":["IBM Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IBM Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5068497719","display_name":"Yotam Perlitz","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yotam Perlitz","raw_affiliation_strings":["IBM Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IBM Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5037214580","display_name":"Roy Bar-Haim","orcid":"https://orcid.org/0000-0001-7365-1662"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Roy Bar-Haim","raw_affiliation_strings":["IBM Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IBM Research","institution_ids":[]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5042370088","display_name":"Lilach Eden","orcid":"https://orcid.org/0000-0001-5244-3813"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lilach Eden","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5024163531","display_name":"Asaf Yehudai","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Asaf Yehudai","raw_affiliation_strings":["IBM Research"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"IBM Research","institution_ids":[]}]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":7.2088,"has_fulltext":true,"cited_by_count":1,"citation_normalized_percentile":{"value":0.96308411,"is_in_top_1_percent":false,"is_in_top_10_percent":true},"cited_by_percentile_year":{"min":94,"max":97},"biblio":{"volume":null,"issue":null,"first_page":"682","last_page":"712"},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10802","display_name":"Judicial and Constitutional Studies","score":0.8955000042915344,"subfield":{"id":"https://openalex.org/subfields/3308","display_name":"Law"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},"topics":[{"id":"https://openalex.org/T10802","display_name":"Judicial and Constitutional Studies","score":0.8955000042915344,"subfield":{"id":"https://openalex.org/subfields/3308","display_name":"Law"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T13710","display_name":"European and International Law Studies","score":0.8741999864578247,"subfield":{"id":"https://openalex.org/subfields/3320","display_name":"Political Science and International Relations"},"field":{"id":"https://openalex.org/fields/33","display_name":"Social Sciences"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}},{"id":"https://openalex.org/T13191","display_name":"Dispute Resolution and Class Actions","score":0.852400004863739,"subfield":{"id":"https://openalex.org/subfields/1408","display_name":"Strategy and Management"},"field":{"id":"https://openalex.org/fields/14","display_name":"Business, Management and Accounting"},"domain":{"id":"https://openalex.org/domains/2","display_name":"Social Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/benchmarking","display_name":"Benchmarking","score":0.9005998373031616},{"id":"https://openalex.org/keywords/ranking","display_name":"Ranking (information retrieval)","score":0.6511713862419128},{"id":"https://openalex.org/keywords/computer-science","display_name":"Computer science","score":0.6135176420211792},{"id":"https://openalex.org/keywords/information-retrieval","display_name":"Information retrieval","score":0.34080952405929565},{"id":"https://openalex.org/keywords/artificial-intelligence","display_name":"Artificial intelligence","score":0.3329855501651764},{"id":"https://openalex.org/keywords/business","display_name":"Business","score":0.06453439593315125}],"concepts":[{"id":"https://openalex.org/C86251818","wikidata":"https://www.wikidata.org/wiki/Q816754","display_name":"Benchmarking","level":2,"score":0.9005998373031616},{"id":"https://openalex.org/C189430467","wikidata":"https://www.wikidata.org/wiki/Q7293293","display_name":"Ranking (information retrieval)","level":2,"score":0.6511713862419128},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6135176420211792},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.34080952405929565},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3329855501651764},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.06453439593315125},{"id":"https://openalex.org/C162853370","wikidata":"https://www.wikidata.org/wiki/Q39809","display_name":"Marketing","level":1,"score":0.0}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.18653/v1/2025.acl-long.34","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.34","pdf_url":"https://aclanthology.org/2025.acl-long.34.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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"}],"best_oa_location":{"id":"doi:10.18653/v1/2025.acl-long.34","is_oa":true,"landing_page_url":"https://doi.org/10.18653/v1/2025.acl-long.34","pdf_url":"https://aclanthology.org/2025.acl-long.34.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 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)","raw_type":"proceedings-article"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W4412886854.pdf","grobid_xml":"https://content.openalex.org/works/W4412886854.grobid-xml"},"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W2118564381","https://openalex.org/W2163901716","https://openalex.org/W2152204162","https://openalex.org/W2739821120","https://openalex.org/W2088097596","https://openalex.org/W2150136235","https://openalex.org/W2140661912","https://openalex.org/W2037724912","https://openalex.org/W2056806613","https://openalex.org/W2153069032"],"abstract_inverted_index":{"Given":[0],"the":[1,18,31,48,51,110,132,139],"rapid":[2],"progress":[3],"of":[4,27,33,50,61,72,114,156],"generative":[5],"AI,":[6],"there":[7],"is":[8,67,135],"a":[9,36,65,70,96,144,153],"pressing":[10],"need":[11],"to":[12,46,80,143],"systematically":[13],"compare":[14],"and":[15,21,25,131,162],"choose":[16],"between":[17],"numerous":[19],"models":[20],"configurations":[22],"available.The":[23],"scale":[24],"versatility":[26],"such":[28,94],"evaluations":[29],"make":[30],"use":[32],"LLMbased":[34],"judges":[35,116],"compelling":[37],"solution":[38],"for":[39],"this":[40,42,86,106],"challenge.Crucially,":[41],"approach":[43],"requires":[44],"first":[45,111],"validate":[47],"quality":[49,134],"LLM":[52,62,115],"judge":[53,66,148,157],"itself.Previous":[54],"work":[55],"has":[56],"focused":[57],"on":[58],"instance-based":[59],"assessment":[60],"judges,":[63],"where":[64],"evaluated":[68],"over":[69,127],"set":[71],"responses,":[73],"or":[74,99],"response":[75],"pairs,":[76],"while":[77],"being":[78],"agnostic":[79],"their":[81,160],"source":[82],"systems.We":[83],"argue":[84],"that":[85],"setting":[87],"overlooks":[88],"critical":[89],"factors":[90],"affecting":[91],"system-level":[92],"ranking,":[93],"as":[95,117],"judge's":[97,133],"positive":[98],"negative":[100],"bias":[101],"towards":[102],"certain":[103],"systems.To":[104],"address":[105],"gap,":[107],"we":[108],"conduct":[109],"large-scale":[112],"study":[113],"system":[118,129,141],"rankers.System":[119],"scores":[120,126],"are":[121],"generated":[122],"by":[123,137],"aggregating":[124],"judgment":[125],"multiple":[128],"outputs,":[130],"assessed":[136],"comparing":[138],"resulting":[140],"ranking":[142],"human-based":[145],"ranking.Beyond":[146],"overall":[147],"assessment,":[149],"our":[150],"analysis":[151],"provides":[152],"fine-grained":[154],"characterization":[155],"behavior,":[158],"including":[159],"decisiveness":[161],"bias.":[163]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-07-29T14:22:42.915294","created_date":"2025-10-10T00:00:00"}