{"id":"https://openalex.org/W7164406311","doi":"https://doi.org/10.48550/arxiv.2606.11209","title":"ProcessThinker: Enhancing Multi-modal Large Language Models Reasoning via Rollout-based Process Reward","display_name":"ProcessThinker: Enhancing Multi-modal Large Language Models Reasoning via Rollout-based Process Reward","publication_year":2026,"publication_date":"2026-04-23","ids":{"openalex":"https://openalex.org/W7164406311","doi":"https://doi.org/10.48550/arxiv.2606.11209"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2606.11209","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.11209","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.2606.11209","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5048975992","display_name":"Jingpei Wu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wu, Jingpei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138425517","display_name":"Xiao Han","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Han, Xiao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5130722573","display_name":"Weixiang Shen","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shen, Weixiang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136785356","display_name":"Boer Zhang","orcid":"https://orcid.org/0000-0001-9752-8239"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Boer","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5138468392","display_name":"Zifeng Ding","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ding, Zifeng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5138424847","display_name":"Volker Tresp","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tresp, Volker","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":0.9937999844551086,"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":0.9937999844551086,"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":0.0010999999940395355,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.0006000000284984708,"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":0.6460999846458435},{"id":"https://openalex.org/keywords/mistake","display_name":"Mistake","score":0.60589998960495},{"id":"https://openalex.org/keywords/pipeline","display_name":"Pipeline (software)","score":0.5763000249862671},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.5012000203132629},{"id":"https://openalex.org/keywords/key","display_name":"Key (lock)","score":0.48159998655319214},{"id":"https://openalex.org/keywords/baseline","display_name":"Baseline (sea)","score":0.4708999991416931},{"id":"https://openalex.org/keywords/verifiable-secret-sharing","display_name":"Verifiable secret sharing","score":0.41690000891685486},{"id":"https://openalex.org/keywords/sample","display_name":"Sample (material)","score":0.3978999853134155},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.37299999594688416}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7766000032424927},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.6460999846458435},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6407999992370605},{"id":"https://openalex.org/C2777179996","wikidata":"https://www.wikidata.org/wiki/Q911222","display_name":"Mistake","level":2,"score":0.60589998960495},{"id":"https://openalex.org/C43521106","wikidata":"https://www.wikidata.org/wiki/Q2165493","display_name":"Pipeline (software)","level":2,"score":0.5763000249862671},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.5012000203132629},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4936000108718872},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.48159998655319214},{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.4708999991416931},{"id":"https://openalex.org/C85847156","wikidata":"https://www.wikidata.org/wiki/Q59015987","display_name":"Verifiable secret sharing","level":3,"score":0.41690000891685486},{"id":"https://openalex.org/C198531522","wikidata":"https://www.wikidata.org/wiki/Q485146","display_name":"Sample (material)","level":2,"score":0.3978999853134155},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.37299999594688416},{"id":"https://openalex.org/C193221554","wikidata":"https://www.wikidata.org/wiki/Q5153664","display_name":"Commonsense reasoning","level":2,"score":0.3474999964237213},{"id":"https://openalex.org/C44291984","wikidata":"https://www.wikidata.org/wiki/Q1074173","display_name":"Question answering","level":2,"score":0.3440000116825104},{"id":"https://openalex.org/C89288958","wikidata":"https://www.wikidata.org/wiki/Q7301504","display_name":"Reasoning system","level":2,"score":0.34150001406669617},{"id":"https://openalex.org/C13662910","wikidata":"https://www.wikidata.org/wiki/Q193139","display_name":"Trajectory","level":2,"score":0.3294000029563904},{"id":"https://openalex.org/C86827895","wikidata":"https://www.wikidata.org/wiki/Q7098582","display_name":"Opportunistic reasoning","level":4,"score":0.3192000091075897},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.31459999084472656},{"id":"https://openalex.org/C2779458634","wikidata":"https://www.wikidata.org/wiki/Q24963715","display_name":"Debiasing","level":2,"score":0.3043999969959259},{"id":"https://openalex.org/C37335422","wikidata":"https://www.wikidata.org/wiki/Q6888134","display_name":"Model-based reasoning","level":3,"score":0.288100004196167},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.2849000096321106},{"id":"https://openalex.org/C166052673","wikidata":"https://www.wikidata.org/wiki/Q83021","display_name":"Empirical evidence","level":2,"score":0.27630001306533813},{"id":"https://openalex.org/C2780586882","wikidata":"https://www.wikidata.org/wiki/Q7520643","display_name":"Simple (philosophy)","level":2,"score":0.2754000127315521},{"id":"https://openalex.org/C43971567","wikidata":"https://www.wikidata.org/wiki/Q3142865","display_name":"Logical reasoning","level":2,"score":0.27079999446868896},{"id":"https://openalex.org/C120936955","wikidata":"https://www.wikidata.org/wiki/Q2155640","display_name":"Empirical research","level":2,"score":0.2680000066757202},{"id":"https://openalex.org/C83725634","wikidata":"https://www.wikidata.org/wiki/Q7268699","display_name":"Qualitative reasoning","level":2,"score":0.2678000032901764},{"id":"https://openalex.org/C62230096","wikidata":"https://www.wikidata.org/wiki/Q275969","display_name":"Crowdsourcing","level":2,"score":0.25060001015663147}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2606.11209","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.11209","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.2606.11209","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2606.11209","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":{"Visual":[0],"question":[1],"answering":[2],"increasingly":[3],"requires":[4,79],"multi-step":[5],"reasoning.":[6,185],"Recent":[7],"post-training":[8,93],"with":[9,121],"reinforcement":[10],"learning":[11],"under":[12],"verifiable":[13],"rewards":[14,99],"(RLVR)":[15],"and":[16,84,126,143,160,194],"Group":[17],"Relative":[18],"Policy":[19],"Optimization":[20],"(GRPO)":[21],"can":[22],"improve":[23],"multimodal":[24],"reasoning,":[25],"but":[26,76],"most":[27],"approaches":[28],"rely":[29],"on":[30],"sparse":[31],"outcome-only":[32],"rewards.":[33],"As":[34],"a":[35,47,68,91,111,122,168,180],"result,":[36],"they":[37],"struggle":[38],"to":[39,66],"tell":[40],"whether":[41],"an":[42,56,102],"incorrect":[43],"answer":[44],"comes":[45],"from":[46,55,59,140],"small":[48],"mistake":[49],"late":[50],"in":[51,183],"the":[52,60,145,152,200],"reasoning":[53,108,162],"or":[54,174],"unhelpful":[57],"trajectory":[58],"start.":[61],"A":[62],"common":[63],"solution":[64],"is":[65],"train":[67],"process":[69,98,129],"reward":[70,125],"model":[71,202],"(PRM)":[72],"for":[73,114,132],"step-level":[74,97],"supervision,":[75],"this":[77],"typically":[78],"large-scale":[80],"high-quality":[81],"chain-of-thought":[82],"annotations":[83],"additional":[85],"training":[86,101],"cost.":[87],"We":[88],"propose":[89],"ProcessThinker,":[90],"practical":[92],"pipeline":[94],"that":[95,141,164],"provides":[96],"without":[100],"explicit":[103],"PRM.":[104],"ProcessThinker":[105,196],"first":[106],"rewrites":[107],"traces":[109],"into":[110],"step-tagged":[112],"format":[113,124],"cold-start":[115],"supervised":[116],"fine-tuning,":[117],"then":[118],"applies":[119],"GRPO":[120],"standard":[123],"our":[127],"rollout-based":[128],"reward.":[130,154],"Concretely,":[131],"each":[133],"intermediate":[134],"step,":[135],"we":[136],"sample":[137],"multiple":[138],"continuations":[139],"step":[142,153],"use":[144],"empirical":[146],"success":[147],"rate":[148],"(final-answer":[149],"verification)":[150],"as":[151],"This":[155],"gives":[156],"dense":[157],"credit":[158],"assignment":[159],"encourages":[161],"steps":[163,178],"more":[165],"reliably":[166],"support":[167],"correct":[169],"conclusion,":[170],"helping":[171],"reduce":[172],"inconsistent":[173],"self-contradictory":[175],"progress":[176],"across":[177],"--":[179],"key":[181],"issue":[182],"logical":[184],"Across":[186],"four":[187],"challenging":[188],"video":[189],"benchmarks":[190],"(Video-MMMU,":[191],"MMVU,":[192],"VideoMathQA,":[193],"LongVideoBench),":[195],"consistently":[197],"improves":[198],"over":[199],"baseline":[201],"Qwen3-VL-8B-Instruct":[203]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-06-12T00:00:00"}