{"id":"https://openalex.org/W7161836690","doi":"https://doi.org/10.48550/arxiv.2605.20164","title":"Not Every Rubric Teaches Equally: Policy-Aware Rubric Rewards for RLVR","display_name":"Not Every Rubric Teaches Equally: Policy-Aware Rubric Rewards for RLVR","publication_year":2026,"publication_date":"2026-05-19","ids":{"openalex":"https://openalex.org/W7161836690","doi":"https://doi.org/10.48550/arxiv.2605.20164"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.20164","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.20164","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,"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":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.20164","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136591831","display_name":"Utkarsh Tyagi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tyagi, Utkarsh","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136515283","display_name":"Xingang Guo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guo, Xingang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5102422296","display_name":"MohammadHossein Rezaei","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Rezaei, MohammadHossein","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133162747","display_name":"Daniel George","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"George, Daniel","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136018321","display_name":"Anas Mahmoud","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mahmoud, Anas","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136588663","display_name":"Jackson Lee","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lee, Jackson","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136606137","display_name":"bing liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Bing","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5079019876","display_name":"Yunzhong He","orcid":"https://orcid.org/0000-0002-5429-5372"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"He, Yunzhong","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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.2632000148296356,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.2632000148296356,"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/T10462","display_name":"Reinforcement Learning in Robotics","score":0.12849999964237213,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.1080000028014183,"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/rubric","display_name":"Rubric","score":0.9724000096321106},{"id":"https://openalex.org/keywords/grading","display_name":"Grading (engineering)","score":0.5576000213623047},{"id":"https://openalex.org/keywords/correctness","display_name":"Correctness","score":0.5232999920845032},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.3343999981880188},{"id":"https://openalex.org/keywords/consistency","display_name":"Consistency (knowledge bases)","score":0.3084000051021576},{"id":"https://openalex.org/keywords/coding","display_name":"Coding (social sciences)","score":0.2955999970436096}],"concepts":[{"id":"https://openalex.org/C111640148","wikidata":"https://www.wikidata.org/wiki/Q847349","display_name":"Rubric","level":2,"score":0.9724000096321106},{"id":"https://openalex.org/C2777286243","wikidata":"https://www.wikidata.org/wiki/Q5591926","display_name":"Grading (engineering)","level":2,"score":0.5576000213623047},{"id":"https://openalex.org/C55439883","wikidata":"https://www.wikidata.org/wiki/Q360812","display_name":"Correctness","level":2,"score":0.5232999920845032},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.42149999737739563},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.3343999981880188},{"id":"https://openalex.org/C2776436953","wikidata":"https://www.wikidata.org/wiki/Q5163215","display_name":"Consistency (knowledge bases)","level":2,"score":0.3084000051021576},{"id":"https://openalex.org/C179518139","wikidata":"https://www.wikidata.org/wiki/Q5140297","display_name":"Coding (social sciences)","level":2,"score":0.2955999970436096},{"id":"https://openalex.org/C2781035248","wikidata":"https://www.wikidata.org/wiki/Q186150","display_name":"Fallacy","level":2,"score":0.2930000126361847},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.28839999437332153},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.27379998564720154},{"id":"https://openalex.org/C145420912","wikidata":"https://www.wikidata.org/wiki/Q853077","display_name":"Mathematics education","level":1,"score":0.2703000009059906},{"id":"https://openalex.org/C73126755","wikidata":"https://www.wikidata.org/wiki/Q7598408","display_name":"Standards-based assessment","level":3,"score":0.26669999957084656},{"id":"https://openalex.org/C127705205","wikidata":"https://www.wikidata.org/wiki/Q5748245","display_name":"Heuristics","level":2,"score":0.2563000023365021}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.20164","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.20164","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,"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":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.20164","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.20164","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,"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":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"},"sustainable_development_goals":[{"display_name":"Peace, Justice and strong institutions","id":"https://metadata.un.org/sdg/16","score":0.46457502245903015}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Reinforcement":[0],"learning":[1],"with":[2,53,89,186],"verifiable":[3],"rewards":[4,29,200],"has":[5],"made":[6],"post-training":[7],"highly":[8],"effective":[9],"when":[10],"correctness":[11],"can":[12,213],"be":[13],"checked":[14],"automatically.":[15],"However,":[16],"many":[17,71],"important":[18,72],"model":[19],"behaviors":[20],"require":[21],"satisfying":[22],"several":[23],"qualitative":[24],"criteria":[25,36,73,81,126],"at":[26],"once.":[27],"Rubric-based":[28],"address":[30],"this":[31,64],"setting":[32],"by":[33],"grading":[34],"prompt-specific":[35],"and":[37,106,154,169,189],"aggregating":[38],"them":[39],"into":[40],"a":[41,49,97],"scalar":[42],"reward.":[43],"Yet":[44],"standard":[45],"static":[46],"aggregations":[47],"conflate":[48],"criterion's":[50],"human-assigned":[51],"importance":[52],"its":[54],"current":[55,216],"usefulness":[56],"as":[57,109],"an":[58],"optimization":[59],"signal.":[60],"We":[61,94],"show":[62],"that":[63,82,102,127],"assumption":[65],"breaks":[66],"down":[67],"in":[68,194,207],"rubric":[69,99,111,167,181,187],"RL:":[70],"are":[74,85],"already":[75],"saturated":[76],"or":[77],"currently":[78,128],"unreachable,":[79],"while":[80,113],"distinguish":[83,203],"rollouts":[84],"not":[86],"necessarily":[87],"those":[88],"the":[90,110,130,134,141,191,208,215],"largest":[91],"human":[92,104],"weights.":[93],"introduce":[95],"POW3R,":[96],"policy-aware":[98],"reward":[100,116,136,168],"framework":[101],"preserves":[103],"weights":[105,117],"category":[107],"balance":[108],"objective":[112],"adapting":[114],"criterion-level":[115],"during":[118],"training.":[119],"POW3R":[120,157],"uses":[121],"rollout-level":[122],"contrast":[123],"to":[124],"emphasize":[125],"separate":[129],"policy's":[131],"outputs,":[132],"making":[133],"GRPO":[135,185],"more":[137],"informative":[138],"without":[139],"changing":[140],"underlying":[142],"evaluation":[143],"target.":[144],"Across":[145],"three":[146],"base":[147],"policies":[148],"on":[149],"two":[150],"datasets":[151],"spanning":[152],"multimodal":[153],"text-only":[155],"settings,":[156],"wins":[158],"$24$":[159],"of":[160,174],"$30$":[161],"base-policy/metric":[162],"comparisons,":[163],"improving":[164],"both":[165],"mean":[166],"strict":[170],"completion":[171],"(the":[172],"fraction":[173],"prompts":[175],"whose":[176],"response":[177],"satisfies":[178],"every":[179],"required":[180],"criterion)":[182],"over":[183],"vanilla":[184],"rewards,":[188],"reaches":[190],"same":[192],"plateau":[193],"$2.5$--$4\\times$":[195],"fewer":[196],"training":[197],"steps.":[198],"Rubric":[199],"should":[201,205],"therefore":[202],"what":[204,212],"matter":[206],"final":[209],"answer":[210],"from":[211],"teach":[214],"policy.":[217]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-05-21T00:00:00"}