{"id":"https://openalex.org/W7154605988","doi":"https://doi.org/10.48550/arxiv.2604.13579","title":"MM-Doc-R1: Training Agents for Long Document Visual Question Answering through Multi-turn Reinforcement Learning","display_name":"MM-Doc-R1: Training Agents for Long Document Visual Question Answering through Multi-turn Reinforcement Learning","publication_year":2026,"publication_date":"2026-04-15","ids":{"openalex":"https://openalex.org/W7154605988","doi":"https://doi.org/10.48550/arxiv.2604.13579"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.13579","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.13579","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.2604.13579","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5025440428","display_name":"Jiaxu Lin","orcid":"https://orcid.org/0009-0005-1777-1690"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Lin, Jiahang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133741696","display_name":"Kai Hu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Hu, Kai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133774689","display_name":"Binghai Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Binghai","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133738589","display_name":"Yuhao Zhou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhou, Yuhao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133756116","display_name":"Zhiheng Xi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Xi, Zhiheng","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133820800","display_name":"Honglin Guo","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Guo, Honglin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133757012","display_name":"Shichun Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Shichun","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133753141","display_name":"Junzhe Wang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Wang, Junzhe","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133810216","display_name":"Shihan Dou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Dou, Shihan","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133729187","display_name":"Enyu Zhou","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhou, Enyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133802983","display_name":"Hang Yan","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yan, Hang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133792805","display_name":"Zhenhua Han","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Han, Zhenhua","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133779211","display_name":"Tao Gui","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Gui, Tao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133800078","display_name":"Qi Zhang","orcid":"https://orcid.org/0009-0001-9927-7936"},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Qi","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5133815587","display_name":"Xuanjing Huang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Huang, Xuanjing","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":0.9937000274658203,"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":0.9937000274658203,"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/T10627","display_name":"Advanced Image and Video Retrieval Techniques","score":0.002199999988079071,"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":0.0008999999845400453,"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/baseline","display_name":"Baseline (sea)","score":0.8578000068664551},{"id":"https://openalex.org/keywords/boosting","display_name":"Boosting (machine learning)","score":0.7623000144958496},{"id":"https://openalex.org/keywords/question-answering","display_name":"Question answering","score":0.7508000135421753},{"id":"https://openalex.org/keywords/reinforcement-learning","display_name":"Reinforcement learning","score":0.6962000131607056},{"id":"https://openalex.org/keywords/workflow","display_name":"Workflow","score":0.6338000297546387},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.5293999910354614}],"concepts":[{"id":"https://openalex.org/C12725497","wikidata":"https://www.wikidata.org/wiki/Q810247","display_name":"Baseline (sea)","level":2,"score":0.8578000068664551},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7692999839782715},{"id":"https://openalex.org/C46686674","wikidata":"https://www.wikidata.org/wiki/Q466303","display_name":"Boosting (machine learning)","level":2,"score":0.7623000144958496},{"id":"https://openalex.org/C44291984","wikidata":"https://www.wikidata.org/wiki/Q1074173","display_name":"Question answering","level":2,"score":0.7508000135421753},{"id":"https://openalex.org/C97541855","wikidata":"https://www.wikidata.org/wiki/Q830687","display_name":"Reinforcement learning","level":2,"score":0.6962000131607056},{"id":"https://openalex.org/C177212765","wikidata":"https://www.wikidata.org/wiki/Q627335","display_name":"Workflow","level":2,"score":0.6338000297546387},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.6128000020980835},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.6085000038146973},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.5293999910354614},{"id":"https://openalex.org/C2164484","wikidata":"https://www.wikidata.org/wiki/Q5170150","display_name":"Core (optical fiber)","level":2,"score":0.4334999918937683},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.36660000681877136},{"id":"https://openalex.org/C23123220","wikidata":"https://www.wikidata.org/wiki/Q816826","display_name":"Information retrieval","level":1,"score":0.2994000017642975},{"id":"https://openalex.org/C155108698","wikidata":"https://www.wikidata.org/wiki/Q1231081","display_name":"Randomized experiment","level":2,"score":0.2687999904155731},{"id":"https://openalex.org/C175154964","wikidata":"https://www.wikidata.org/wiki/Q380077","display_name":"Task analysis","level":3,"score":0.2565000057220459},{"id":"https://openalex.org/C2777211547","wikidata":"https://www.wikidata.org/wiki/Q17141490","display_name":"Training (meteorology)","level":2,"score":0.2508000135421753}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.13579","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.13579","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.2604.13579","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.13579","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":[{"display_name":"Peace, Justice and strong institutions","score":0.4494571387767792,"id":"https://metadata.un.org/sdg/16"}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Conventional":[0],"Retrieval-Augmented":[1],"Generation":[2],"(RAG)":[3],"systems":[4],"often":[5],"struggle":[6],"with":[7,169,173],"complex":[8],"multi-hop":[9],"queries":[10],"over":[11,163],"long":[12,33],"documents":[13],"due":[14],"to":[15,31,120,137],"their":[16,90],"single-pass":[17],"retrieval.":[18],"We":[19],"introduce":[20],"MM-Doc-R1,":[21],"a":[22,99,126],"novel":[23,185],"framework":[24,183],"that":[25,76,141,151],"employs":[26],"an":[27],"agentic,":[28],"vision-aware":[29],"workflow":[30],"address":[32],"document":[34],"visual":[35,195],"question":[36,196],"answering":[37],"through":[38],"iterative":[39],"information":[40,47],"discovery":[41],"and":[42,129,171,184],"synthesis.":[43],"To":[44],"incentivize":[45],"the":[46,80,87,116,147,178,190],"seeking":[48],"capabilities":[49],"of":[50,106,180],"our":[51,134,181],"agents,":[52,135],"we":[53],"propose":[54],"Similarity-based":[55],"Policy":[56],"Optimization":[57],"(SPO),":[58],"addressing":[59],"baseline":[60,92,102,119],"estimation":[61,93],"bias":[62],"in":[63,77,188],"existing":[64],"multi-turn":[65,78],"reinforcement":[66],"learning":[67,131],"(RL)":[68],"algorithms":[69],"like":[70],"GRPO.":[71,143],"Our":[72,144],"core":[73],"insight":[74],"is":[75],"RL,":[79],"more":[81,88,100,127],"semantically":[82],"similar":[83],"two":[84],"trajectories":[85],"are,":[86],"accurate":[89,130],"shared":[91],"becomes.":[94],"Leveraging":[95],"this,":[96],"SPO":[97,159],"calculates":[98],"precise":[101],"by":[103,156,167],"similarity-weighted":[104],"averaging":[105],"rewards":[107],"across":[108],"multiple":[109],"trajectories,":[110],"unlike":[111],"GRPO":[112],"which":[113],"inappropriately":[114],"applies":[115],"initial":[117],"state's":[118],"all":[121],"intermediate":[122],"states.":[123],"This":[124],"provides":[125],"stable":[128],"signal":[132],"for":[133,192],"leading":[136],"superior":[138,161],"training":[139,186],"performance":[140,162],"surpasses":[142],"experiments":[145],"on":[146],"MMLongbench-Doc":[148],"benchmark":[149],"show":[150],"MM-Doc-R1":[152],"outperforms":[153],"previous":[154],"baselines":[155],"10.4%.":[157],"Furthermore,":[158],"demonstrates":[160],"GRPO,":[164],"boosting":[165],"results":[166,176],"5.0%":[168],"Qwen3-8B":[170],"6.1%":[172],"Qwen3-4B.":[174],"These":[175],"highlight":[177],"effectiveness":[179],"integrated":[182],"algorithm":[187],"advancing":[189],"state-of-the-art":[191],"complex,":[193],"long-document":[194],"answering.":[197]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-04-17T00:00:00"}