{"id":"https://openalex.org/W7161041800","doi":"https://doi.org/10.48550/arxiv.2605.12034","title":"Boosting Omni-Modal Language Models: Staged Post-Training with Visually Debiased Evaluation","display_name":"Boosting Omni-Modal Language Models: Staged Post-Training with Visually Debiased Evaluation","publication_year":2026,"publication_date":"2026-05-12","ids":{"openalex":"https://openalex.org/W7161041800","doi":"https://doi.org/10.48550/arxiv.2605.12034"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2605.12034","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.12034","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.2605.12034","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5136030105","display_name":"Che Liu","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Liu, Che","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136035551","display_name":"Lichao Ma","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ma, Lichao","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136019239","display_name":"Xiangyu Tony Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Xiangyu Tony","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136060292","display_name":"Yuxin Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Yuxin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136028808","display_name":"Haoyang Zhang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhang, Haoyang","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5136066999","display_name":"Xuerui Yang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Yang, Xuerui","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5136026317","display_name":"Fei Tian","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Tian, Fei","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":0.8995000123977661,"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":0.8995000123977661,"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/T11309","display_name":"Music and Audio Processing","score":0.01730000041425228,"subfield":{"id":"https://openalex.org/subfields/1711","display_name":"Signal Processing"},"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/T10775","display_name":"Generative Adversarial Networks and Image Synthesis","score":0.010300000198185444,"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"}}],"keywords":[{"id":"https://openalex.org/keywords/boosting","display_name":"Boosting (machine learning)","score":0.7448999881744385},{"id":"https://openalex.org/keywords/benchmark","display_name":"Benchmark (surveying)","score":0.6159999966621399},{"id":"https://openalex.org/keywords/audit","display_name":"Audit","score":0.510200023651123},{"id":"https://openalex.org/keywords/language-model","display_name":"Language model","score":0.37860000133514404},{"id":"https://openalex.org/keywords/gradient-boosting","display_name":"Gradient boosting","score":0.31189998984336853}],"concepts":[{"id":"https://openalex.org/C46686674","wikidata":"https://www.wikidata.org/wiki/Q466303","display_name":"Boosting (machine learning)","level":2,"score":0.7448999881744385},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.6992999911308289},{"id":"https://openalex.org/C185798385","wikidata":"https://www.wikidata.org/wiki/Q1161707","display_name":"Benchmark (surveying)","level":2,"score":0.6159999966621399},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5273000001907349},{"id":"https://openalex.org/C199521495","wikidata":"https://www.wikidata.org/wiki/Q181487","display_name":"Audit","level":2,"score":0.510200023651123},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.4083999991416931},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.37860000133514404},{"id":"https://openalex.org/C31972630","wikidata":"https://www.wikidata.org/wiki/Q844240","display_name":"Computer vision","level":1,"score":0.32260000705718994},{"id":"https://openalex.org/C70153297","wikidata":"https://www.wikidata.org/wiki/Q5591907","display_name":"Gradient boosting","level":3,"score":0.31189998984336853},{"id":"https://openalex.org/C106131492","wikidata":"https://www.wikidata.org/wiki/Q3072260","display_name":"Filter (signal processing)","level":2,"score":0.29820001125335693},{"id":"https://openalex.org/C51632099","wikidata":"https://www.wikidata.org/wiki/Q3985153","display_name":"Training set","level":2,"score":0.27000001072883606},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.2696000039577484},{"id":"https://openalex.org/C2986089797","wikidata":"https://www.wikidata.org/wiki/Q6501338","display_name":"Visual attention","level":3,"score":0.2596000134944916}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2605.12034","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.12034","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.2605.12034","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2605.12034","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":[{"display_name":"Quality Education","score":0.7719137072563171,"id":"https://metadata.un.org/sdg/4"}],"awards":[],"funders":[],"has_content":{"pdf":false,"grobid_xml":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Omni-modal":[0],"language":[1],"models":[2,178],"are":[3],"intended":[4],"to":[5,25,167],"jointly":[6],"understand":[7],"audio,":[8],"visual":[9,20,36,172],"inputs,":[10],"and":[11,43,65,110,120,129,147,174],"language,":[12],"but":[13],"benchmark":[14,133],"gains":[15],"can":[16,179],"be":[17],"inflated":[18],"when":[19,69,169],"evidence":[21,41],"alone":[22],"is":[23,71,165],"enough":[24],"answer":[26],"a":[27,48,81,98,155],"query.":[28],"We":[29,53],"study":[30],"whether":[31],"current":[32],"omni-modal":[33,56,157,163,177],"benchmarks":[34,57],"separate":[35],"shortcuts":[37],"from":[38,89,181],"genuine":[39],"audio-visual-language":[40],"integration,":[42],"how":[44],"post-training":[45,100,183],"behaves":[46],"under":[47],"visually":[49,62],"debiased":[50],"evaluation":[51,83,170],"setting.":[52],"audit":[54],"nine":[55],"with":[58,85,184],"visual-only":[59],"probing,":[60],"remove":[61],"solvable":[63],"queries,":[64],"retain":[66],"full":[67],"subsets":[68],"filtering":[70],"undefined":[72],"or":[73],"would":[74],"make":[75],"comparisons":[76],"unstable.":[77],"This":[78],"yields":[79],"OmniClean,":[80,94],"cleaned":[82],"view":[84],"8,551":[86],"retained":[87],"queries":[88],"16,968":[90],"audited":[91],"queries.":[92],"On":[93],"we":[95],"evaluate":[96],"OmniBoost,":[97],"three-stage":[99],"recipe":[101],"based":[102],"on":[103,112,137],"Qwen2.5-Omni-3B:":[104],"mixed":[105],"bi-modal":[106,116],"SFT,":[107],"mixed-modality":[108],"RLVR,":[109],"SFT":[111,117,136],"self-distilled":[113,138,185],"data.":[114],"Balanced":[115],"gives":[118],"limited":[119],"uneven":[121],"gains,":[122],"RLVR":[123],"provides":[124],"the":[125,132,140],"first":[126],"broad":[127],"improvement,":[128],"self-distillation":[130],"reshapes":[131],"profile.":[134],"After":[135],"data,":[139],"3B":[141],"model":[142],"reaches":[143],"performance":[144],"comparable":[145],"to,":[146],"in":[148],"aggregate":[149],"slightly":[150],"above,":[151],"Qwen3-Omni-30B-A3B-Instruct":[152],"without":[153],"using":[154],"stronger":[156],"teacher.":[158],"These":[159],"results":[160],"show":[161],"that":[162,175],"progress":[164],"easier":[166],"interpret":[168],"controls":[171],"leakage,":[173],"small":[176],"benefit":[180],"staged":[182],"omni-query":[186],"supervision.":[187],"Project":[188],"page:":[189],"https://cheliu-computation.github.io/omni/":[190]},"counts_by_year":[],"updated_date":"2026-07-01T06:00:48.157686","created_date":"2026-05-14T00:00:00"}