{"id":"https://openalex.org/W7127948103","doi":"https://doi.org/10.48550/arxiv.2602.05125","title":"Rethinking Rubric Generation for Improving LLM Judge and Reward Modeling for Open-ended Tasks","display_name":"Rethinking Rubric Generation for Improving LLM Judge and Reward Modeling for Open-ended Tasks","publication_year":2026,"publication_date":"2026-02-04","ids":{"openalex":"https://openalex.org/W7127948103","doi":"https://doi.org/10.48550/arxiv.2602.05125"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2602.05125","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2602.05125","pdf_url":"https://arxiv.org/pdf/2602.05125","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2602.05125","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5113222165","display_name":"William F. Shen","orcid":"https://orcid.org/0009-0007-2327-1697","observed_orcids":["https://orcid.org/0009-0007-2327-1697"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shen, William F.","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5040821950","display_name":"Xinchi Qiu","orcid":"https://orcid.org/0000-0002-7654-4448","observed_orcids":["https://orcid.org/0000-0002-7654-4448"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Qiu, Xinchi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5047340661","display_name":"Chenxi Whitehouse","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Whitehouse, Chenxi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5001337463","display_name":"Lisa Alazraki","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Alazraki, Lisa","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5030450290","display_name":"Shashwat Goel","orcid":null,"observed_orcids":[]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Goel, Shashwat","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102984496","display_name":"Francesco Barbieri","orcid":"https://orcid.org/0000-0002-2144-8440","observed_orcids":["https://orcid.org/0000-0002-2144-8440"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Barbieri, Francesco","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5041285498","display_name":"Timon Willi","orcid":"https://orcid.org/0000-0003-4405-5700","observed_orcids":["https://orcid.org/0000-0003-4405-5700"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Willi, Timon","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5020784139","display_name":"Akhil Mathur","orcid":"https://orcid.org/0000-0002-1475-3017","observed_orcids":["https://orcid.org/0000-0002-1475-3017"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mathur, Akhil","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5066833626","display_name":"Ilias Leontiadis A","orcid":"https://orcid.org/0000-0001-5581-5803","observed_orcids":["https://orcid.org/0000-0001-5581-5803"]},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Leontiadis, Ilias","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":null,"has_fulltext":false,"cited_by_count":0,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T10028","display_name":"Topic Modeling","score":0.5307999849319458,"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":0.5307999849319458,"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":0.10119999945163727,"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/T10286","display_name":"Information Retrieval and Search Behavior","score":0.026200000196695328,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/rubric-generation","display_name":"rubric generation","score":0.9890000224113464},{"id":"https://openalex.org/keywords/llm-as-a-judge","display_name":"LLM-as-a-judge","score":0.8429999947547913},{"id":"https://openalex.org/keywords/reward-modeling","display_name":"reward modeling","score":0.8889999985694885},{"id":"https://openalex.org/keywords/reinforcement-fine-tuning","display_name":"reinforcement fine-tuning","score":0.9049999713897705},{"id":"https://openalex.org/keywords/recursive-decomposition","display_name":"recursive decomposition","score":0.5260000228881836},{"id":"https://openalex.org/keywords/human-preference-alignment","display_name":"human preference alignment","score":0.5189999938011169},{"id":"https://openalex.org/keywords/open-ended-tasks","display_name":"open-ended tasks","score":0.6859999895095825}],"concepts":[{"id":"https://openalex.org/C111640148","wikidata":"https://www.wikidata.org/wiki/Q847349","display_name":"Rubric","level":2,"score":0.9309999942779541},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6761999726295471},{"id":"https://openalex.org/C183115368","wikidata":"https://www.wikidata.org/wiki/Q856577","display_name":"Weighting","level":2,"score":0.4982999861240387},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.4878000020980835},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4700999855995178},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.4607999920845032},{"id":"https://openalex.org/C97931131","wikidata":"https://www.wikidata.org/wiki/Q5282087","display_name":"Discriminative model","level":2,"score":0.4577000141143799},{"id":"https://openalex.org/C46686674","wikidata":"https://www.wikidata.org/wiki/Q466303","display_name":"Boosting (machine learning)","level":2,"score":0.42820000648498535},{"id":"https://openalex.org/C2776493592","wikidata":"https://www.wikidata.org/wiki/Q5158717","display_name":"Succinctness","level":2,"score":0.31049999594688416},{"id":"https://openalex.org/C184898388","wikidata":"https://www.wikidata.org/wiki/Q1435712","display_name":"Pairwise comparison","level":2,"score":0.3050999939441681},{"id":"https://openalex.org/C2776436953","wikidata":"https://www.wikidata.org/wiki/Q5163215","display_name":"Consistency (knowledge bases)","level":2,"score":0.301800012588501},{"id":"https://openalex.org/C77618280","wikidata":"https://www.wikidata.org/wiki/Q1155772","display_name":"Scheme (mathematics)","level":2,"score":0.2985000014305115},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.2928999960422516},{"id":"https://openalex.org/C127705205","wikidata":"https://www.wikidata.org/wiki/Q5748245","display_name":"Heuristics","level":2,"score":0.28439998626708984},{"id":"https://openalex.org/C2781162219","wikidata":"https://www.wikidata.org/wiki/Q26250693","display_name":"Replicate","level":2,"score":0.271699994802475},{"id":"https://openalex.org/C2781067378","wikidata":"https://www.wikidata.org/wiki/Q17027399","display_name":"Interpretability","level":2,"score":0.2703000009059906},{"id":"https://openalex.org/C117251300","wikidata":"https://www.wikidata.org/wiki/Q1849855","display_name":"Parametric statistics","level":2,"score":0.2694999873638153},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.26499998569488525},{"id":"https://openalex.org/C2777146004","wikidata":"https://www.wikidata.org/wiki/Q14949826","display_name":"CLARITY","level":2,"score":0.2635999917984009},{"id":"https://openalex.org/C2780586970","wikidata":"https://www.wikidata.org/wiki/Q1357284","display_name":"Popularity","level":2,"score":0.2628999948501587},{"id":"https://openalex.org/C2781137444","wikidata":"https://www.wikidata.org/wiki/Q237105","display_name":"Sharpening","level":2,"score":0.2578999996185303},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.2533999979496002},{"id":"https://openalex.org/C11012388","wikidata":"https://www.wikidata.org/wiki/Q170658","display_name":"Creativity","level":2,"score":0.25290000438690186},{"id":"https://openalex.org/C2781249084","wikidata":"https://www.wikidata.org/wiki/Q908656","display_name":"Preference","level":2,"score":0.2515000104904175}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2602.05125","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2602.05125","pdf_url":"https://arxiv.org/pdf/2602.05125","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2602.05125","doi":"https://doi.org/10.48550/arxiv.2602.05125","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.05125","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2602.05125","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2602.05125","pdf_url":"https://arxiv.org/pdf/2602.05125","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"sustainable_development_goals_aurora":[{"id":"https://metadata.un.org/sdg/10","score":0.6663415431976318,"display_name":"Reduced inequalities"}],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":["https://openalex.org/W7167225266","https://openalex.org/W7163068252","https://openalex.org/W7163013473","https://openalex.org/W7134860561","https://openalex.org/W7202353701","https://openalex.org/W7202185450","https://openalex.org/W7204964375","https://openalex.org/W7160826640","https://openalex.org/W4417112261","https://openalex.org/W4414490998"],"abstract_inverted_index":{"Recently,":[0],"rubrics":[1,36,79],"have":[2,17],"been":[3,18],"used":[4,158],"to":[5,22,34,152,180,196],"guide":[6],"LLM":[7,214],"judges":[8],"in":[9,147,219],"capturing":[10],"subjective,":[11],"nuanced,":[12],"multi-dimensional":[13],"human":[14],"preferences,":[15],"and":[16,45,55,97,100,117,128,136,141,171,183,198,210,216],"extended":[19],"from":[20],"evaluation":[21,127],"reward":[23,161,177,217],"signals":[24],"for":[25,67,138,163,188,213],"reinforcement":[26],"fine-tuning":[27],"(RFT).":[28],"However,":[29],"rubric":[30,68,111,190,205],"generation":[31],"remains":[32],"hard":[33],"control:":[35],"often":[37],"lack":[38],"coverage,":[39],"conflate":[40],"dimensions,":[41],"misalign":[42],"preference":[43],"direction,":[44],"contain":[46],"redundant":[47,98],"or":[48],"highly":[49,107],"correlated":[50,108],"criteria,":[51,83,109],"degrading":[52],"judge":[53],"accuracy":[54,133],"producing":[56],"suboptimal":[57],"rewards":[58],"during":[59],"RFT.":[60],"We":[61],"propose":[62],"RRD,":[63],"a":[64,72,101,208],"principled":[65],"framework":[66],"refinement":[69,206],"built":[70],"on":[71,134,155,165],"recursive":[73,204],"decompose-filter":[74],"cycle.":[75],"RRD":[76,120,202],"decomposes":[77],"coarse":[78],"into":[80],"fine-grained,":[81],"discriminative":[82],"expanding":[84],"coverage":[85],"while":[86],"sharpening":[87],"separation":[88],"between":[89],"responses.":[90],"A":[91],"complementary":[92],"filtering":[93],"mechanism":[94],"removes":[95],"misaligned":[96],"rubrics,":[99],"correlation-aware":[102],"weighting":[103],"scheme":[104],"prevents":[105],"over-representing":[106],"yielding":[110],"sets":[112],"that":[113,194],"are":[114],"informative,":[115],"comprehensive,":[116],"non-redundant.":[118],"Empirically,":[119],"delivers":[121],"large,":[122],"consistent":[123],"gains":[124,193],"across":[125],"both":[126,139],"training:":[129],"it":[130,167],"improves":[131],"preference-judgment":[132],"JudgeBench":[135],"PPE":[137],"GPT-4o":[140],"Llama3.1-405B":[142],"judges,":[143],"achieving":[144],"top":[145],"performance":[146],"all":[148],"settings":[149],"with":[150,192],"up":[151,179],"+17.7":[153],"points":[154],"JudgeBench.":[156],"When":[157],"as":[159,207],"the":[160],"source":[162],"RFT":[164],"WildChat,":[166],"yields":[168],"substantially":[169],"stronger":[170],"more":[172],"stable":[173],"learning":[174],"signals,":[175],"boosting":[176],"by":[178],"160%":[181],"(Qwen3-4B)":[182],"60%":[184],"(Llama3.1-8B)":[185],"versus":[186],"10-20%":[187],"prior":[189],"baselines,":[191],"transfer":[195],"HealthBench-Hard":[197],"BiGGen":[199],"Bench.":[200],"Overall,":[201],"establishes":[203],"scalable":[209],"interpretable":[211],"foundation":[212],"judging":[215],"modeling":[218],"open-ended":[220],"domains.":[221]},"counts_by_year":[],"updated_date":"2026-10-28T10:33:24.094632","created_date":"2026-02-07T00:00:00"}