{"id":"https://openalex.org/W7160901314","doi":"https://doi.org/10.48550/arxiv.2605.09269","title":"DeltaRubric: Generative Multimodal Reward Modeling via Joint Planning and Verification","display_name":"DeltaRubric: Generative Multimodal Reward Modeling via Joint Planning and Verification","publication_year":2026,"publication_date":"2026-05-10","ids":{"openalex":"https://openalex.org/W7160901314","doi":"https://doi.org/10.48550/arxiv.2605.09269"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.09269","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.09269","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":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.2605.09269","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5135964515","display_name":"Rui Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Rui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135983684","display_name":"Dian Yu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yu, Dian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135922406","display_name":"Zhenwen Liang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liang, Zhenwen","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135934973","display_name":"Yucheng Shi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shi, Yucheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135964431","display_name":"Tong Zheng","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zheng, Tong","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136001547","display_name":"Runpeng Dai","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dai, Runpeng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135954739","display_name":"Haitao Mi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Mi, Haitao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5135913760","display_name":"Pratap Tokekar","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tokekar, Pratap","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5135928747","display_name":"Leoweiliang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Leoweiliang","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":0.9585999846458435,"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":0.9585999846458435,"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":0.010599999688565731,"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/T12026","display_name":"Explainable Artificial Intelligence (XAI)","score":0.006200000178068876,"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/process","display_name":"Process (computing)","score":0.5095999836921692},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.49050000309944153},{"id":"https://openalex.org/keywords/generative-grammar","display_name":"Generative grammar","score":0.44589999318122864},{"id":"https://openalex.org/keywords/structured-prediction","display_name":"Structured prediction","score":0.423799991607666},{"id":"https://openalex.org/keywords/joint","display_name":"Joint (building)","score":0.41670000553131104},{"id":"https://openalex.org/keywords/prior-probability","display_name":"Prior probability","score":0.366100013256073},{"id":"https://openalex.org/keywords/generative-model","display_name":"Generative model","score":0.3571000099182129},{"id":"https://openalex.org/keywords/set","display_name":"Set (abstract data type)","score":0.3472000062465668}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8242999911308289},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6223999857902527},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.5386999845504761},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.5095999836921692},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.49050000309944153},{"id":"https://openalex.org/C39890363","wikidata":"https://www.wikidata.org/wiki/Q36108","display_name":"Generative grammar","level":2,"score":0.44589999318122864},{"id":"https://openalex.org/C22367795","wikidata":"https://www.wikidata.org/wiki/Q7625208","display_name":"Structured prediction","level":2,"score":0.423799991607666},{"id":"https://openalex.org/C18555067","wikidata":"https://www.wikidata.org/wiki/Q8375051","display_name":"Joint (building)","level":2,"score":0.41670000553131104},{"id":"https://openalex.org/C177769412","wikidata":"https://www.wikidata.org/wiki/Q278090","display_name":"Prior probability","level":3,"score":0.366100013256073},{"id":"https://openalex.org/C167966045","wikidata":"https://www.wikidata.org/wiki/Q5532625","display_name":"Generative model","level":3,"score":0.3571000099182129},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.3472000062465668},{"id":"https://openalex.org/C2780801425","wikidata":"https://www.wikidata.org/wiki/Q5164392","display_name":"Construct (python library)","level":2,"score":0.32739999890327454},{"id":"https://openalex.org/C157657479","wikidata":"https://www.wikidata.org/wiki/Q2367247","display_name":"Closed captioning","level":3,"score":0.3070000112056732},{"id":"https://openalex.org/C111640148","wikidata":"https://www.wikidata.org/wiki/Q847349","display_name":"Rubric","level":2,"score":0.305400013923645},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.301800012588501},{"id":"https://openalex.org/C165064840","wikidata":"https://www.wikidata.org/wiki/Q1321061","display_name":"Matching (statistics)","level":2,"score":0.29789999127388},{"id":"https://openalex.org/C147764199","wikidata":"https://www.wikidata.org/wiki/Q6865248","display_name":"Minification","level":2,"score":0.29670000076293945},{"id":"https://openalex.org/C85847156","wikidata":"https://www.wikidata.org/wiki/Q59015987","display_name":"Verifiable secret sharing","level":3,"score":0.2892000079154968},{"id":"https://openalex.org/C2780910867","wikidata":"https://www.wikidata.org/wiki/Q1952416","display_name":"Multimodality","level":2,"score":0.28630000352859497},{"id":"https://openalex.org/C42058472","wikidata":"https://www.wikidata.org/wiki/Q810214","display_name":"Base (topology)","level":2,"score":0.2849999964237213},{"id":"https://openalex.org/C126042441","wikidata":"https://www.wikidata.org/wiki/Q1324888","display_name":"Frame (networking)","level":2,"score":0.2847000062465668},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.2777999937534332},{"id":"https://openalex.org/C2776214188","wikidata":"https://www.wikidata.org/wiki/Q408386","display_name":"Inference","level":2,"score":0.27219998836517334}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.09269","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.09269","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":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.2605.09269","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.09269","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":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":{"Aligning":[0],"Multimodal":[1],"Large":[2],"Language":[3],"Models":[4],"(MLLMs)":[5],"requires":[6,61],"reliable":[7,195],"reward":[8,199],"models,":[9,154],"yet":[10],"existing":[11],"single-step":[12],"evaluators":[13],"can":[14],"suffer":[15],"from":[16],"lazy":[17],"judging,":[18],"exploiting":[19],"language":[20],"priors":[21],"over":[22],"fine-grained":[23],"visual":[24,46,57],"verification.":[25],"While":[26],"rubric-based":[27],"evaluation":[28,60,83,187],"mitigates":[29],"these":[30,118],"biases":[31],"in":[32,94],"text-only":[33],"settings,":[34],"extending":[35],"it":[36,116,164],"to":[37,126,193],"multimodal":[38,81,198],"tasks":[39],"is":[40],"bottlenecked":[41],"by":[42,170],"the":[43,103,122,128],"complexity":[44],"of":[45],"reasoning.":[47],"The":[48,182],"critical":[49],"differences":[50],"between":[51],"responses":[52],"often":[53],"depend":[54],"on":[55,148],"instance-specific":[56,108],"details.":[58],"Robust":[59],"dynamically":[62],"synthesizing":[63],"rubrics":[64],"that":[65,79,185],"isolate":[66],"spatial":[67],"and":[68,124,144,151,173,196],"factual":[69],"discrepancies.":[70],"To":[71],"address":[72],"this,":[73],"we":[74],"introduce":[75],"$\\textbf{DeltaRubric}$,":[76],"an":[77],"approach":[78],"reformulates":[80],"preference":[82],"as":[84,99,135],"a":[85,89,100,106,113,136],"plan-and-execute":[86],"process":[87],"within":[88],"single":[90],"MLLM.":[91],"DeltaRubric":[92,134,155],"operates":[93],"two":[95],"steps:":[96],"acting":[97],"first":[98],"$\\textit{Disagreement":[101],"Planner}$,":[102],"model":[104,167],"generates":[105],"neutral,":[107],"verification":[109,145],"checklist.":[110],"Transitioning":[111],"into":[112,188],"$\\textit{Checklist":[114],"Verifier}$,":[115],"executes":[117],"self-generated":[119],"checks":[120],"against":[121],"image":[123],"question":[125],"produce":[127],"final":[129],"grounded":[130],"judgment.":[131],"We":[132],"formulate":[133],"multi-role":[137],"reinforcement":[138],"learning":[139],"problem,":[140],"jointly":[141],"optimizing":[142],"planning":[143],"capabilities.":[146],"Validated":[147],"Qwen3-VL":[149],"4B":[150],"8B":[152],"Instruct":[153],"achieves":[156],"solid":[157],"empirical":[158],"gains.":[159],"For":[160],"instance,":[161],"On":[162],"VL-RewardBench,":[163],"improves":[165],"base":[166],"overall":[168],"accuracy":[169],"$\\textbf{+22.6}$":[171],"(4B)":[172],"$\\textbf{+18.8}$":[174],"(8B)":[175],"points,":[176],"largely":[177],"outperforming":[178],"standard":[179],"no-rubric":[180],"baselines.":[181],"results":[183],"demonstrate":[184],"decomposing":[186],"structured,":[189],"verifiable":[190],"steps":[191],"leads":[192],"more":[194],"generalizable":[197],"modeling.":[200]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-13T00:00:00"}