{"id":"https://openalex.org/W7162113940","doi":"https://doi.org/10.48550/arxiv.2605.21801","title":"Why Semantic Entropy Fails: Geometry-Aware and Calibrated Uncertainty for Policy Optimization","display_name":"Why Semantic Entropy Fails: Geometry-Aware and Calibrated Uncertainty for Policy Optimization","publication_year":2026,"publication_date":"2026-05-20","ids":{"openalex":"https://openalex.org/W7162113940","doi":"https://doi.org/10.48550/arxiv.2605.21801"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.21801","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.21801","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2605.21801","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136793833","display_name":"Zheyuan Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Zheyuan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136810481","display_name":"Kaiwen Shi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shi, Kaiwen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136760708","display_name":"Han Bao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Bao, Han","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136812180","display_name":"Zehong Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Zehong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136801806","display_name":"Tianyi Ma","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ma, Tianyi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136776781","display_name":"Yanfang Ye","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ye, Yanfang","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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.1889999955892563,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T11714","display_name":"Multimodal Machine Learning Applications","score":0.1889999955892563,"subfield":{"id":"https://openalex.org/subfields/1707","display_name":"Computer Vision and Pattern Recognition"},"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/T10028","display_name":"Topic Modeling","score":0.1518000066280365,"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":0.12319999933242798,"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/leverage","display_name":"Leverage (statistics)","score":0.7210000157356262},{"id":"https://openalex.org/keywords/calibration","display_name":"Calibration","score":0.5315999984741211},{"id":"https://openalex.org/keywords/perspective","display_name":"Perspective (graphical)","score":0.48500001430511475},{"id":"https://openalex.org/keywords/robust-optimization","display_name":"Robust optimization","score":0.46129998564720154},{"id":"https://openalex.org/keywords/variance","display_name":"Variance (accounting)","score":0.4408999979496002},{"id":"https://openalex.org/keywords/entropy","display_name":"Entropy (arrow of time)","score":0.43560001254081726},{"id":"https://openalex.org/keywords/scalability","display_name":"Scalability","score":0.4284999966621399},{"id":"https://openalex.org/keywords/optimization-problem","display_name":"Optimization problem","score":0.42829999327659607}],"concepts":[{"id":"https://openalex.org/C153083717","wikidata":"https://www.wikidata.org/wiki/Q6535263","display_name":"Leverage (statistics)","level":2,"score":0.7210000157356262},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6736000180244446},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5428000092506409},{"id":"https://openalex.org/C165838908","wikidata":"https://www.wikidata.org/wiki/Q736777","display_name":"Calibration","level":2,"score":0.5315999984741211},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5239999890327454},{"id":"https://openalex.org/C12713177","wikidata":"https://www.wikidata.org/wiki/Q1900281","display_name":"Perspective (graphical)","level":2,"score":0.48500001430511475},{"id":"https://openalex.org/C193254401","wikidata":"https://www.wikidata.org/wiki/Q2160088","display_name":"Robust optimization","level":2,"score":0.46129998564720154},{"id":"https://openalex.org/C196083921","wikidata":"https://www.wikidata.org/wiki/Q7915758","display_name":"Variance (accounting)","level":2,"score":0.4408999979496002},{"id":"https://openalex.org/C106301342","wikidata":"https://www.wikidata.org/wiki/Q4117933","display_name":"Entropy (arrow of time)","level":2,"score":0.43560001254081726},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.4284999966621399},{"id":"https://openalex.org/C137836250","wikidata":"https://www.wikidata.org/wiki/Q984063","display_name":"Optimization problem","level":2,"score":0.42829999327659607},{"id":"https://openalex.org/C2987595161","wikidata":"https://www.wikidata.org/wiki/Q141495","display_name":"Optimization algorithm","level":2,"score":0.3499999940395355},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.3334999978542328},{"id":"https://openalex.org/C63479239","wikidata":"https://www.wikidata.org/wiki/Q7353546","display_name":"Robustness (evolution)","level":3,"score":0.33149999380111694},{"id":"https://openalex.org/C2778049539","wikidata":"https://www.wikidata.org/wiki/Q17002908","display_name":"Bayesian optimization","level":2,"score":0.32600000500679016},{"id":"https://openalex.org/C99498987","wikidata":"https://www.wikidata.org/wiki/Q2210247","display_name":"Noise (video)","level":3,"score":0.325300008058548},{"id":"https://openalex.org/C2779843651","wikidata":"https://www.wikidata.org/wiki/Q7390335","display_name":"SIGNAL (programming language)","level":2,"score":0.3179999887943268},{"id":"https://openalex.org/C9679016","wikidata":"https://www.wikidata.org/wiki/Q1417473","display_name":"Principle of maximum entropy","level":2,"score":0.3127000033855438},{"id":"https://openalex.org/C32230216","wikidata":"https://www.wikidata.org/wiki/Q7882499","display_name":"Uncertainty quantification","level":2,"score":0.2680000066757202},{"id":"https://openalex.org/C137209882","wikidata":"https://www.wikidata.org/wiki/Q1403517","display_name":"Measurement uncertainty","level":2,"score":0.26409998536109924},{"id":"https://openalex.org/C124101348","wikidata":"https://www.wikidata.org/wiki/Q172491","display_name":"Data mining","level":1,"score":0.2531999945640564}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.21801","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.21801","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":"cc-by","license_id":"https://openalex.org/licenses/cc-by","version":null,"is_accepted":false,"is_published":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2605.21801","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.21801","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":null,"issn":null,"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":"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"},"sustainable_development_goals":[],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Post-training":[0],"has":[1],"become":[2],"central":[3],"to":[4,26,40,67,130,137],"improving":[5],"reasoning":[6],"and":[7,54,82,86,94,109,157],"alignment":[8],"in":[9,56],"large":[10],"language":[11],"models,":[12],"where":[13],"critic-free":[14],"models":[15],"enable":[16],"scalable":[17],"learning":[18,87,141],"from":[19,29],"model-generated":[20],"outputs":[21],"but":[22],"lack":[23],"principled":[24,72,179],"mechanisms":[25,79],"distinguish":[27],"informative":[28],"noisy":[30],"signals.":[31],"Recent":[32],"approaches":[33],"leverage":[34],"response-level":[35],"measures":[36,129],"as":[37,46,78],"uncertainty":[38,76,139,169],"signals":[39,77,170],"regulate":[41],"group-based":[42],"optimization":[43,60,175],"methods":[44],"such":[45],"GRPO.":[47],"Yet":[48],"their":[49],"empirical":[50,93],"success":[51],"remains":[52],"unstable":[53],"unclear":[55],"how":[57],"they":[58],"influence":[59],"dynamics.":[61],"In":[62],"this":[63,115],"paper,":[64],"we":[65,97,117],"provide,":[66],"our":[68,150],"knowledge,":[69],"the":[70,165],"first":[71],"formulation":[73],"that":[74,149,171],"interprets":[75],"for":[80,181],"characterizing":[81],"regulating":[83],"gradient":[84,155],"variance":[85],"signal":[88,142],"quality.":[89],"Based":[90],"on":[91,145],"both":[92],"theoretical":[95],"analysis,":[96,116],"identify":[98],"two":[99],"critical":[100],"gaps":[101],"of":[102,167],"current":[103],"entropy-based":[104],"estimators:":[105],"The":[106,110],"anisotropic":[107],"gap":[108],"calibration":[111,136],"gap.":[112],"Motivated":[113],"by":[114],"propose":[118],"Geometric-aware":[119],"Calibrated":[120],"Policy":[121],"Optimization":[122],"(GCPO),":[123],"a":[124,178],"novel":[125],"framework":[126],"integrating":[127],"geometry-aware":[128],"capture":[131],"semantic":[132],"disagreement":[133],"with":[134,140,174],"reward-based":[135],"align":[138],"strength.":[143],"Experiments":[144],"multiple":[146],"benchmarks":[147],"show":[148],"approach":[151],"more":[152],"faithfully":[153],"tracks":[154],"variability":[156],"consistently":[158],"improves":[159],"post-training":[160],"performance.":[161],"Our":[162],"results":[163],"highlight":[164],"importance":[166],"designing":[168],"are":[172],"aligned":[173],"dynamics,":[176],"offering":[177],"perspective":[180],"robust":[182],"post-training.":[183]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-23T00:00:00"}