{"id":"https://openalex.org/W7163674141","doi":"https://doi.org/10.48550/arxiv.2606.05174","title":"Improving Heart-Focused Medical Question Answering in LLMs via Variance-Aware Rubric Rewards with GRPO","display_name":"Improving Heart-Focused Medical Question Answering in LLMs via Variance-Aware Rubric Rewards with GRPO","publication_year":2026,"publication_date":"2026-04-17","ids":{"openalex":"https://openalex.org/W7163674141","doi":"https://doi.org/10.48550/arxiv.2606.05174"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.05174","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.05174","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.2606.05174","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5137927350","display_name":"Arash Ahmadi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ahmadi, Arash","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5137961380","display_name":"Parisa Masnadi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Khiabani, Parisa Masnadi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5128194497","display_name":"Sarah Sharif","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Sharif, Sarah","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5117812694","display_name":"Charles Nicholson","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Nicholson, Charles","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5120765849","display_name":"David Ebert","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ebert, David","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5095926817","display_name":"Mike Banad","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Banad, Mike","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":0.43939998745918274,"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":0.43939998745918274,"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/T13702","display_name":"Machine Learning in Healthcare","score":0.17800000309944153,"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":0.15760000050067902,"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/rubric","display_name":"Rubric","score":0.92330002784729},{"id":"https://openalex.org/keywords/inference","display_name":"Inference","score":0.5856000185012817},{"id":"https://openalex.org/keywords/health-care","display_name":"Health care","score":0.3792000114917755},{"id":"https://openalex.org/keywords/causal-inference","display_name":"Causal inference","score":0.32260000705718994},{"id":"https://openalex.org/keywords/ranking","display_name":"Ranking (information retrieval)","score":0.3181000053882599},{"id":"https://openalex.org/keywords/construct","display_name":"Construct (python library)","score":0.31060001254081726},{"id":"https://openalex.org/keywords/variety","display_name":"Variety (cybernetics)","score":0.305400013923645},{"id":"https://openalex.org/keywords/rank","display_name":"Rank (graph theory)","score":0.2904999852180481}],"concepts":[{"id":"https://openalex.org/C111640148","wikidata":"https://www.wikidata.org/wiki/Q847349","display_name":"Rubric","level":2,"score":0.92330002784729},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.5856000185012817},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5715000033378601},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.3797999918460846},{"id":"https://openalex.org/C160735492","wikidata":"https://www.wikidata.org/wiki/Q31207","display_name":"Health care","level":2,"score":0.3792000114917755},{"id":"https://openalex.org/C2522767166","wikidata":"https://www.wikidata.org/wiki/Q2374463","display_name":"Data science","level":1,"score":0.335099995136261},{"id":"https://openalex.org/C15744967","wikidata":"https://www.wikidata.org/wiki/Q9418","display_name":"Psychology","level":0,"score":0.3244999945163727},{"id":"https://openalex.org/C158600405","wikidata":"https://www.wikidata.org/wiki/Q5054566","display_name":"Causal inference","level":2,"score":0.32260000705718994},{"id":"https://openalex.org/C539667460","wikidata":"https://www.wikidata.org/wiki/Q2414942","display_name":"Management science","level":1,"score":0.3181999921798706},{"id":"https://openalex.org/C189430467","wikidata":"https://www.wikidata.org/wiki/Q7293293","display_name":"Ranking (information retrieval)","level":2,"score":0.3181000053882599},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.31060001254081726},{"id":"https://openalex.org/C136197465","wikidata":"https://www.wikidata.org/wiki/Q1729295","display_name":"Variety (cybernetics)","level":2,"score":0.305400013923645},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.3052999973297119},{"id":"https://openalex.org/C164226766","wikidata":"https://www.wikidata.org/wiki/Q7293202","display_name":"Rank (graph theory)","level":2,"score":0.2904999852180481},{"id":"https://openalex.org/C58546491","wikidata":"https://www.wikidata.org/wiki/Q1150207","display_name":"Competitive advantage","level":2,"score":0.28780001401901245},{"id":"https://openalex.org/C204434341","wikidata":"https://www.wikidata.org/wiki/Q357789","display_name":"Adjudication","level":2,"score":0.2851000130176544},{"id":"https://openalex.org/C56739046","wikidata":"https://www.wikidata.org/wiki/Q192060","display_name":"Knowledge management","level":1,"score":0.27970001101493835},{"id":"https://openalex.org/C44291984","wikidata":"https://www.wikidata.org/wiki/Q1074173","display_name":"Question answering","level":2,"score":0.2757999897003174},{"id":"https://openalex.org/C42058472","wikidata":"https://www.wikidata.org/wiki/Q810214","display_name":"Base (topology)","level":2,"score":0.27230000495910645},{"id":"https://openalex.org/C2781035248","wikidata":"https://www.wikidata.org/wiki/Q186150","display_name":"Fallacy","level":2,"score":0.2653000056743622},{"id":"https://openalex.org/C162307627","wikidata":"https://www.wikidata.org/wiki/Q204833","display_name":"Enhanced Data Rates for GSM Evolution","level":2,"score":0.26499998569488525},{"id":"https://openalex.org/C48372109","wikidata":"https://www.wikidata.org/wiki/Q3913","display_name":"Binary number","level":2,"score":0.2639999985694885},{"id":"https://openalex.org/C75684735","wikidata":"https://www.wikidata.org/wiki/Q858810","display_name":"Big data","level":2,"score":0.25999999046325684}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.05174","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.05174","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.2606.05174","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.05174","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":[],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Large":[0],"Language":[1],"Models":[2],"(LLMs)":[3],"have":[4],"shown":[5],"strong":[6],"promise":[7],"in":[8,15,199],"healthcare":[9],"applications.":[10],"Yet":[11],"deploying":[12],"general-purpose":[13],"models":[14,44],"real-world":[16],"settings":[17],"remains":[18],"difficult":[19,132],"due":[20],"to":[21,50,133,158,164,167,203,205],"data":[22],"privacy":[23],"constraints,":[24],"inference":[25],"costs,":[26],"and":[27,90,104,131,136,160],"limited":[28],"suitability":[29],"for":[30,65,125,193],"edge":[31],"or":[32],"on-device":[33],"use.":[34],"These":[35],"challenges":[36],"motivate":[37],"the":[38,87,168],"development":[39],"of":[40,94,148],"smaller,":[41],"more":[42,138],"efficient":[43],"that":[45,85,127,184],"require":[46],"robust":[47],"post-training":[48,66],"strategies":[49,93],"ensure":[51],"reliable":[52],"medical":[53,70,196],"reasoning.":[54],"In":[55],"this":[56],"work,":[57],"we":[58],"investigate":[59],"Group":[60],"Relative":[61],"Policy":[62],"Optimization":[63],"(GRPO)":[64],"LLMs":[67],"on":[68],"heart-focused":[69,195],"question":[71,197],"answering":[72,198],"with":[73,109,175,201],"rubric-based":[74,187,207],"supervision":[75],"derived":[76,114],"from":[77,115,156,162],"RaR-Medicine.":[78],"We":[79],"propose":[80],"a":[81,144,190],"Variance-Aware":[82],"Reward":[83],"Framework":[84],"extends":[86],"Explicit":[88],"Aggregation":[89,92],"Implicit":[91],"Rubrics":[95],"as":[96],"Rewards":[97],"by":[98],"replacing":[99],"weighted":[100],"binary":[101],"criterion":[102],"aggregation":[103],"single":[105],"overall":[106],"Likert-style":[107],"scoring":[108],"continuous":[110],"analytical":[111],"reward":[112],"functions":[113],"criterion-level":[116],"rubric":[117],"outcomes.":[118],"This":[119],"formulation":[120],"provides":[121],"richer":[122],"optimization":[123],"signals":[124],"feedback":[126],"is":[128],"sparse,":[129],"multi-criteria,":[130],"verify":[134],"automatically,":[135],"enables":[137],"stable":[139],"on-policy":[140],"reinforcement":[141],"learning.":[142],"On":[143],"held-out":[145],"heart-related":[146],"subset":[147],"HealthBench,":[149],"our":[150],"best":[151],"GRPO":[152],"variant":[153],"improves":[154],"accuracy":[155],"0.362":[157],"0.502":[159],"F1":[161],"0.532":[163],"0.668":[165],"relative":[166],"Qwen3-14B":[169],"base":[170],"model,":[171],"while":[172],"remaining":[173],"competitive":[174],"GPT-OSS-120B":[176],"(0.508":[177],"accuracy,":[178],"0.674":[179],"F1).":[180],"Our":[181],"findings":[182],"show":[183],"carefully":[185],"designed":[186],"rewards":[188],"provide":[189],"practical":[191],"strategy":[192],"improving":[194],"LLMs,":[200],"potential":[202],"extend":[204],"other":[206],"tasks.":[208]},"counts_by_year":[],"updated_date":"2026-08-01T09:00:35.917206","created_date":"2026-06-06T00:00:00"}