{"id":"https://openalex.org/W7127424698","doi":"https://doi.org/10.48550/arxiv.2602.00505","title":"Sparse Shortcuts: Facilitating Efficient Fusion in Multimodal Large Language Models","display_name":"Sparse Shortcuts: Facilitating Efficient Fusion in Multimodal Large Language Models","publication_year":2026,"publication_date":"2026-01-31","ids":{"openalex":"https://openalex.org/W7127424698","doi":"https://doi.org/10.48550/arxiv.2602.00505"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2602.00505","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2602.00505","pdf_url":"https://arxiv.org/pdf/2602.00505","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2602.00505","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5075392699","display_name":"Jingrui Zhang","orcid":"https://orcid.org/0000-0001-7732-8459","observed_orcids":["https://orcid.org/0000-0001-7732-8459"]},"institutions":[{"id":"https://openalex.org/I4388482657","display_name":"Shenzhen MSU-BIT University","ror":"https://ror.org/02q963474","country_code":"CN","type":"education","lineage":["https://openalex.org/I4388482657"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhang, Jingrui","raw_affiliation_strings":["Guangdong-Hong Kong-Macao Joint Laboratory for Emotion Intelligence and Pervasive Computing , Artificial Intelligence Research Institute , Shenzhen MSU-BIT University , Shenzhen , 518172 , Guangdong , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Guangdong-Hong Kong-Macao Joint Laboratory for Emotion Intelligence and Pervasive Computing , Artificial Intelligence Research Institute , Shenzhen MSU-BIT University , Shenzhen , 518172 , Guangdong , China","institution_ids":["https://openalex.org/I4388482657"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5134133655","display_name":"Feng Liang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4388482657","display_name":"Shenzhen MSU-BIT University","ror":"https://ror.org/02q963474","country_code":"CN","type":"education","lineage":["https://openalex.org/I4388482657"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Liang, Feng","raw_affiliation_strings":["Guangdong-Hong Kong-Macao Joint Laboratory for Emotion Intelligence and Pervasive Computing , Artificial Intelligence Research Institute , Shenzhen MSU-BIT University , Shenzhen , 518172 , Guangdong , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Guangdong-Hong Kong-Macao Joint Laboratory for Emotion Intelligence and Pervasive Computing , Artificial Intelligence Research Institute , Shenzhen MSU-BIT University , Shenzhen , 518172 , Guangdong , China","institution_ids":["https://openalex.org/I4388482657"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5124874260","display_name":"Yong Zhang","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4388482657","display_name":"Shenzhen MSU-BIT University","ror":"https://ror.org/02q963474","country_code":"CN","type":"education","lineage":["https://openalex.org/I4388482657"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zhang, Yong","raw_affiliation_strings":["Guangdong-Hong Kong-Macao Joint Laboratory for Emotion Intelligence and Pervasive Computing , Artificial Intelligence Research Institute , Shenzhen MSU-BIT University , Shenzhen , 518172 , Guangdong , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Guangdong-Hong Kong-Macao Joint Laboratory for Emotion Intelligence and Pervasive Computing , Artificial Intelligence Research Institute , Shenzhen MSU-BIT University , Shenzhen , 518172 , Guangdong , China","institution_ids":["https://openalex.org/I4388482657"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5100391896","display_name":"Wei William Wang","orcid":"https://orcid.org/0000-0002-1717-5785","observed_orcids":["https://orcid.org/0000-0002-1717-5785","https://orcid.org/0000-0001-8676-1190","https://orcid.org/0000-0002-0307-0296","https://orcid.org/0000-0002-5987-738X","https://orcid.org/0000-0002-1102-1777","https://orcid.org/0000-0002-9325-3677","https://orcid.org/0000-0001-8980-2680","https://orcid.org/0000-0002-8598-0831","https://orcid.org/0000-0001-5873-2384"]},"institutions":[{"id":"https://openalex.org/I4388482657","display_name":"Shenzhen MSU-BIT University","ror":"https://ror.org/02q963474","country_code":"CN","type":"education","lineage":["https://openalex.org/I4388482657"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Wang, Wei","raw_affiliation_strings":["Guangdong-Hong Kong-Macao Joint Laboratory for Emotion Intelligence and Pervasive Computing , Artificial Intelligence Research Institute , Shenzhen MSU-BIT University , Shenzhen , 518172 , Guangdong , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Guangdong-Hong Kong-Macao Joint Laboratory for Emotion Intelligence and Pervasive Computing , Artificial Intelligence Research Institute , Shenzhen MSU-BIT University , Shenzhen , 518172 , Guangdong , China","institution_ids":["https://openalex.org/I4388482657"]}]},{"author_position":"middle","author":{"id":"https://openalex.org/A5018498977","display_name":"Runhao Zeng","orcid":"https://orcid.org/0000-0001-8694-4245","observed_orcids":["https://orcid.org/0000-0001-8694-4245"]},"institutions":[{"id":"https://openalex.org/I4388482657","display_name":"Shenzhen MSU-BIT University","ror":"https://ror.org/02q963474","country_code":"CN","type":"education","lineage":["https://openalex.org/I4388482657"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Zeng, Runhao","raw_affiliation_strings":["Guangdong-Hong Kong-Macao Joint Laboratory for Emotion Intelligence and Pervasive Computing , Artificial Intelligence Research Institute , Shenzhen MSU-BIT University , Shenzhen , 518172 , Guangdong , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Guangdong-Hong Kong-Macao Joint Laboratory for Emotion Intelligence and Pervasive Computing , Artificial Intelligence Research Institute , Shenzhen MSU-BIT University , Shenzhen , 518172 , Guangdong , China","institution_ids":["https://openalex.org/I4388482657"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5123694260","display_name":"Xiping Hu","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I4388482657","display_name":"Shenzhen MSU-BIT University","ror":"https://ror.org/02q963474","country_code":"CN","type":"education","lineage":["https://openalex.org/I4388482657"]}],"countries":["CN"],"is_corresponding":false,"raw_author_name":"Hu, Xiping","raw_affiliation_strings":["Guangdong-Hong Kong-Macao Joint Laboratory for Emotion Intelligence and Pervasive Computing , Artificial Intelligence Research Institute , Shenzhen MSU-BIT University , Shenzhen , 518172 , Guangdong , China"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Guangdong-Hong Kong-Macao Joint Laboratory for Emotion Intelligence and Pervasive Computing , Artificial Intelligence Research Institute , Shenzhen MSU-BIT University , Shenzhen , 518172 , Guangdong , China","institution_ids":["https://openalex.org/I4388482657"]}]}],"institutions":[],"countries_distinct_count":1,"institutions_distinct_count":1,"corresponding_author_ids":[],"corresponding_institution_ids":["https://openalex.org/I4388482657"],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":1,"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":1.0,"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":1.0,"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/T10036","display_name":"Advanced Neural Network Applications","score":0.0,"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/T11307","display_name":"Domain Adaptation and Few-Shot Learning","score":0.0,"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/multimodal-large-language-models","display_name":"multimodal large language models","score":0.9929999709129333},{"id":"https://openalex.org/keywords/cross-modal-fusion","display_name":"cross-modal fusion","score":0.9620000123977661},{"id":"https://openalex.org/keywords/visual-feature-fusion","display_name":"visual feature fusion","score":0.6570000052452087},{"id":"https://openalex.org/keywords/vision-language-models","display_name":"vision-language models","score":0.7129999995231628},{"id":"https://openalex.org/keywords/low-level-visual-features","display_name":"low-level visual features","score":0.7329999804496765}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.8312000036239624},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.5795000195503235},{"id":"https://openalex.org/C2780767217","wikidata":"https://www.wikidata.org/wiki/Q5532421","display_name":"Generality","level":2,"score":0.5631999969482422},{"id":"https://openalex.org/C195324797","wikidata":"https://www.wikidata.org/wiki/Q33742","display_name":"Natural language","level":2,"score":0.5404999852180481},{"id":"https://openalex.org/C48044578","wikidata":"https://www.wikidata.org/wiki/Q727490","display_name":"Scalability","level":2,"score":0.5131000280380249},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.47290000319480896},{"id":"https://openalex.org/C98045186","wikidata":"https://www.wikidata.org/wiki/Q205663","display_name":"Process (computing)","level":2,"score":0.453900009393692},{"id":"https://openalex.org/C2779439875","wikidata":"https://www.wikidata.org/wiki/Q1078276","display_name":"Natural language understanding","level":3,"score":0.4422000050544739},{"id":"https://openalex.org/C2777508537","wikidata":"https://www.wikidata.org/wiki/Q7936620","display_name":"Visual reasoning","level":2,"score":0.42289999127388},{"id":"https://openalex.org/C119857082","wikidata":"https://www.wikidata.org/wiki/Q2539","display_name":"Machine learning","level":1,"score":0.41499999165534973},{"id":"https://openalex.org/C2779903281","wikidata":"https://www.wikidata.org/wiki/Q6888026","display_name":"Modalities","level":2,"score":0.40130001306533813},{"id":"https://openalex.org/C2776401178","wikidata":"https://www.wikidata.org/wiki/Q12050496","display_name":"Feature (linguistics)","level":2,"score":0.4007999897003174},{"id":"https://openalex.org/C137293760","wikidata":"https://www.wikidata.org/wiki/Q3621696","display_name":"Language model","level":2,"score":0.3991999924182892},{"id":"https://openalex.org/C49937458","wikidata":"https://www.wikidata.org/wiki/Q2599292","display_name":"Probabilistic logic","level":2,"score":0.3912000060081482},{"id":"https://openalex.org/C192209626","wikidata":"https://www.wikidata.org/wiki/Q190909","display_name":"Focus (optics)","level":2,"score":0.3797000050544739},{"id":"https://openalex.org/C66024118","wikidata":"https://www.wikidata.org/wiki/Q1122506","display_name":"Computational model","level":2,"score":0.36079999804496765},{"id":"https://openalex.org/C204321447","wikidata":"https://www.wikidata.org/wiki/Q30642","display_name":"Natural language processing","level":1,"score":0.35839998722076416},{"id":"https://openalex.org/C184337299","wikidata":"https://www.wikidata.org/wiki/Q1437428","display_name":"Semantics (computer science)","level":2,"score":0.3546999990940094},{"id":"https://openalex.org/C183322885","wikidata":"https://www.wikidata.org/wiki/Q17007702","display_name":"Context model","level":3,"score":0.3336000144481659},{"id":"https://openalex.org/C118505674","wikidata":"https://www.wikidata.org/wiki/Q42586063","display_name":"Encoder","level":2,"score":0.3264000117778778},{"id":"https://openalex.org/C36464697","wikidata":"https://www.wikidata.org/wiki/Q451553","display_name":"Visualization","level":2,"score":0.302700012922287},{"id":"https://openalex.org/C2780878386","wikidata":"https://www.wikidata.org/wiki/Q1659648","display_name":"Visual language","level":2,"score":0.2919999957084656},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.2865000069141388},{"id":"https://openalex.org/C158525013","wikidata":"https://www.wikidata.org/wiki/Q2593739","display_name":"Fusion","level":2,"score":0.2849000096321106},{"id":"https://openalex.org/C4554734","wikidata":"https://www.wikidata.org/wiki/Q593744","display_name":"Knowledge base","level":2,"score":0.2800999879837036},{"id":"https://openalex.org/C173414695","wikidata":"https://www.wikidata.org/wiki/Q5510276","display_name":"Fusion mechanism","level":4,"score":0.2793000042438507},{"id":"https://openalex.org/C33954974","wikidata":"https://www.wikidata.org/wiki/Q486494","display_name":"Sensor fusion","level":2,"score":0.27799999713897705},{"id":"https://openalex.org/C174252522","wikidata":"https://www.wikidata.org/wiki/Q3816772","display_name":"Natural language user interface","level":3,"score":0.2667999863624573},{"id":"https://openalex.org/C188198153","wikidata":"https://www.wikidata.org/wiki/Q1613840","display_name":"Limiting","level":2,"score":0.2605000138282776},{"id":"https://openalex.org/C155092808","wikidata":"https://www.wikidata.org/wiki/Q182557","display_name":"Computational linguistics","level":2,"score":0.2556000053882599}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2602.00505","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2602.00505","pdf_url":"https://arxiv.org/pdf/2602.00505","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2602.00505","doi":"https://doi.org/10.48550/arxiv.2602.00505","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2602.00505","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2602.00505","doi":null,"is_oa":true,"landing_page_url":"http://arxiv.org/abs/2602.00505","pdf_url":"https://arxiv.org/pdf/2602.00505","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"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":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7127424698.pdf","grobid_xml":"https://content.openalex.org/works/W7127424698.grobid-xml"},"referenced_works_count":14,"referenced_works":["https://openalex.org/W3094502228","https://openalex.org/W4229005866","https://openalex.org/W4322718191","https://openalex.org/W4327810158","https://openalex.org/W4362515116","https://openalex.org/W4382491206","https://openalex.org/W4384918448","https://openalex.org/W4385473806","https://openalex.org/W4389523832","https://openalex.org/W4407049645","https://openalex.org/W4409269797","https://openalex.org/W4409348271","https://openalex.org/W4410704295","https://openalex.org/W4413849607"],"related_works":["https://openalex.org/W4403882354","https://openalex.org/W7220391629","https://openalex.org/W4416036089","https://openalex.org/W4405033094","https://openalex.org/W4413144552","https://openalex.org/W4415962928","https://openalex.org/W7124195832","https://openalex.org/W7155049678","https://openalex.org/W4415428181","https://openalex.org/W7141388862"],"abstract_inverted_index":{"With":[0],"the":[1,58,75,86,111,115,121,151,160,164,172,184,192],"remarkable":[2],"success":[3],"of":[4,89,126,153,194],"large":[5,16],"language":[6,11,17,40,59,166],"models":[7,18,41],"(LLMs)":[8],"in":[9,23,80,180],"natural":[10],"understanding":[12],"and":[13,82,114,123,168,202],"generation,":[14],"multimodal":[15,198],"(MLLMs)":[19],"have":[20],"rapidly":[21],"advanced":[22],"their":[24],"ability":[25,88],"to":[26,52],"process":[27],"data":[28],"across":[29,196],"multiple":[30,130],"modalities.":[31],"While":[32],"most":[33],"existing":[34],"efforts":[35],"focus":[36],"on":[37],"scaling":[38],"up":[39],"or":[42],"constructing":[43],"higher-quality":[44],"training":[45],"data,":[46],"limited":[47],"attention":[48],"has":[49],"been":[50],"paid":[51],"effectively":[53],"integrating":[54],"cross-modal":[55,101,112],"knowledge":[56],"into":[57],"space.":[60],"In":[61],"vision-language":[62],"models,":[63],"for":[64,104,183,204],"instance,":[65],"aligning":[66],"modalities":[67],"using":[68],"only":[69],"high-level":[70],"visual":[71,127,154],"features":[72,128,155],"often":[73],"discards":[74],"rich":[76],"semantic":[77,134],"information":[78],"present":[79],"mid-":[81],"low-level":[83],"features,":[84],"limiting":[85],"model's":[87],"cross-modality":[90],"understanding.":[91],"To":[92],"address":[93],"this":[94],"issue,":[95],"we":[96],"propose":[97],"SparseCut,":[98],"a":[99],"general":[100],"fusion":[102,135,147,152],"architecture":[103],"MLLMs,":[105],"introducing":[106],"sparse":[107],"shortcut":[108,118],"connections":[109,119],"between":[110],"encoder":[113],"LLM.":[116,185],"These":[117],"enable":[120],"efficient":[122,144],"hierarchical":[124],"integration":[125],"at":[129],"levels,":[131],"facilitating":[132],"richer":[133],"without":[136],"increasing":[137],"computational":[138,181],"overhead.":[139],"We":[140],"further":[141],"introduce":[142],"an":[143,178],"multi-grained":[145],"feature":[146],"module,":[148],"which":[149],"performs":[150],"before":[156],"routing":[157],"them":[158],"through":[159],"shortcuts.":[161],"This":[162],"preserves":[163],"original":[165],"context":[167],"does":[169],"not":[170],"increase":[171,179],"overall":[173],"input":[174],"length,":[175],"thereby":[176],"avoiding":[177],"complexity":[182],"Experiments":[186],"demonstrate":[187],"that":[188],"SparseCut":[189],"significantly":[190],"enhances":[191],"performance":[193],"MLLMs":[195],"various":[197],"benchmarks":[199],"with":[200],"generality":[201],"scalability":[203],"different":[205],"base":[206],"LLMs.":[207]},"counts_by_year":[{"year":2026,"cited_by_count":1}],"updated_date":"2026-11-10T11:39:43.021295","created_date":"2026-02-04T00:00:00"}